Merge branch 'main' of ssh://gitea.rmontanana.es:6422/rmontanana/Platform

Reviewed-on: #6

.gitignore

@@ -41,3 +41,7 @@ puml/**
 *.dot
 diagrams/html/**
 diagrams/latex/**
+.cache
+vcpkg_installed
+.claude/settings.local.json
+CMakeUserPresets.json

.gitmodules

@@ -1,21 +0,0 @@
-[submodule "lib/catch2"]
-	path = lib/catch2
-	url = https://github.com/catchorg/Catch2.git
-[submodule "lib/argparse"]
-	path = lib/argparse
-	url = https://github.com/p-ranav/argparse
-[submodule "lib/json"]
-	path = lib/json
-	url = https://github.com/nlohmann/json
-[submodule "lib/libxlsxwriter"]
-	path = lib/libxlsxwriter
-	url = https://github.com/jmcnamara/libxlsxwriter.git
-[submodule "lib/folding"]
-	path = lib/folding
-	url = https://github.com/rmontanana/folding
-[submodule "lib/Files"]
-	path = lib/Files
-	url = https://github.com/rmontanana/ArffFiles
-[submodule "lib/mdlp"]
-	path = lib/mdlp
-	url = https://github.com/rmontanana/mdlp

CHANGELOG.md

@@ -0,0 +1,93 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+### Changed
+- **BREAKING**: Migrated dependency management from vcpkg to Conan
+- Updated build system to use Conan toolchain files instead of vcpkg
+- Updated `make init` command to use `conan install` instead of `vcpkg install`
+- Modified CMakeLists.txt to use Conan's find_package mechanism
+- Updated documentation in CLAUDE.md to reflect Conan usage
+
+### Added
+- `conanfile.py` - Conan recipe for dependency management with all required dependencies
+- CMakeUserPresets.json (generated by Conan)
+- Support for Conan build profiles (Release/Debug)
+
+### Removed
+- `vcpkg.json` - vcpkg manifest file
+- `vcpkg-configuration.json` - vcpkg registry configuration
+- vcpkg toolchain dependency in build system
+
+### Notes
+- The migration maintains compatibility with existing make targets and workflow
+- All dependencies now managed through Conan package manager
+
+## [1.1.0] - 2025-07-02
+
+### Added
+- **AdaBoost Implementation**: Complete multi-class SAMME AdaBoost classifier with optimization
+  - Optimized AdaBoostPredict with 100 estimators as default
+  - Enhanced predictProbaSample functionality
+  - Full predict_proba support for probabilistic predictions
+- **Decision Tree Classifier**: New base classifier implementation with comprehensive tests
+- **XA1DE Model Family**: Extended Averaged One-Dependence Estimators
+  - XA1DE, XBAODE, XSPODE variants with threading support
+  - Complete integration with memory optimization
+  - Prior probability computation in prediction
+- **Wilcoxon Statistical Test**: Statistical significance testing for model comparison
+- **Folder Management**: Enhanced file organization with folder parameter support across tools
+  - Added folder parameter to b_best, b_grid, b_main, and b_manage
+- **vcpkg Integration**: Package management system integration (now migrated to Conan)
+
+### Enhanced
+- **Grid Search System**: Complete refactoring with MPI parallelization
+  - Grid experiment functionality with conditional result saving
+  - Fixed smoothing problems and dataset ordering
+  - Enhanced reporting and summary generation
+- **Excel Reporting**: Advanced Excel export capabilities
+  - ReportExcelCompared class for side-by-side result comparison
+  - Enhanced formatting with colors and fixed headers
+  - Automatic file opening after generation
+- **Results Management**: Comprehensive result handling and validation
+  - JSON schema validation for result format integrity
+  - Improved console reporting with classification reports
+  - Pagination support for large result sets
+- **Statistical Analysis**: Enhanced statistical testing and reporting
+  - AUC (Area Under Curve) computation and reporting
+  - Confusion matrix generation and visualization
+  - Classification reports with color coding
+
+### Performance Improvements
+- Optimized AdaBoost training and prediction algorithms
+- Enhanced memory management in XA1DE implementations
+- Improved discretization algorithms with MDLP integration
+- Faster ROC-AUC computation for binary classification problems
+
+### Developer Experience
+- **Testing Framework**: Comprehensive test suite with Catch2
+- **Build System**: Streamlined CMake configuration with dependency management
+- **Documentation**: Enhanced project documentation and build instructions
+- **Code Quality**: Refactored codebase with improved error handling and logging
+
+### Bug Fixes
+- Fixed predict_proba implementations across multiple classifiers
+- Resolved grid search dataset ordering issues  
+- Fixed Excel report formatting and column width problems
+- Corrected time output formatting in various tools
+- Fixed memory leaks and stability issues in model implementations
+
+## [1.0.0] - 2024-01-09
+
+### Initial Release
+- **Core Framework**: Machine learning experimentation platform for Bayesian Networks
+- **Basic Classifiers**: Initial set of Bayesian network classifiers
+- **Experiment Management**: Basic experiment orchestration and result storage
+- **Dataset Support**: ARFF file format support with discretization
+- **Build System**: CMake-based build system with external library integration
+- **Command Line Tools**: Initial versions of b_main, b_best, b_list utilities

CLAUDE.md

@@ -0,0 +1,139 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+Platform is a C++ machine learning framework for running experiments with Bayesian Networks and other classifiers. It supports both research-focused experimental classifiers and production-ready models through a unified interface.
+
+## Build System
+
+The project uses CMake with Make as the primary build system:
+
+- **Release build**: `make release` (creates `build_Release/` directory)
+- **Debug build**: `make debug` (creates `build_Debug/` directory with testing and coverage enabled)
+- **Install binaries**: `make install` (copies executables to `~/bin` by default)
+- **Clean project**: `make clean` (removes build directories)
+- **Initialize dependencies**: `make init` (runs conan install for both Release and Debug)
+
+### Testing
+
+- **Run tests**: `make test` (builds debug version and runs all tests)
+- **Coverage report**: `make coverage` (runs tests and generates coverage with gcovr)
+- **Single test with options**: `make test opt="-s"` (verbose) or `make test opt="-c='Test Name'"` (specific test)
+
+### Build Targets
+
+Main executables (built from `src/commands/`):
+- `b_main`: Main experiment runner
+- `b_grid`: Grid search over hyperparameters 
+- `b_best`: Best results analysis and comparison
+- `b_list`: Dataset listing and properties
+- `b_manage`: Results management interface
+- `b_results`: Results processing
+
+## Dependencies
+
+The project uses Conan for package management with these key dependencies:
+- **libtorch**: PyTorch C++ backend for tensor operations
+- **nlohmann_json**: JSON processing
+- **catch2**: Unit testing framework
+- **cli11**: Command-line argument parsing (replacement for argparse)
+
+Custom dependencies (not available in ConanCenter):
+- **fimdlp**: MDLP discretization library (needs manual integration)
+- **folding**: Cross-validation utilities (needs manual integration)
+- **arff-files**: ARFF dataset file handling (needs manual integration)
+
+External dependencies (managed separately):
+- **BayesNet**: Core Bayesian network classifiers (from `../lib/`)
+- **PyClassifiers**: Python classifier wrappers (from `../lib/`)
+- **MPI**: Message Passing Interface for parallel processing
+- **Boost**: Python integration and utilities
+
+**Note**: Some dependencies (fimdlp, folding, arff-files) are not available in ConanCenter and need to be:
+- Built as custom Conan packages, or
+- Integrated using CMake FetchContent, or
+- Built separately and found via find_package
+
+## Architecture
+
+### Core Components
+
+**Experiment Framework** (`src/main/`):
+- `Experiment.cpp/h`: Main experiment orchestration
+- `Models.cpp/h`: Classifier factory and registration system
+- `Scores.cpp/h`: Performance metrics calculation
+- `HyperParameters.cpp/h`: Parameter management
+- `ArgumentsExperiment.cpp/h`: Command-line argument handling
+
+**Data Handling** (`src/common/`):
+- `Dataset.cpp/h`: Individual dataset representation
+- `Datasets.cpp/h`: Dataset collection management
+- `Discretization.cpp/h`: Data discretization utilities
+
+**Classifiers** (`src/experimental_clfs/`):
+- `AdaBoost.cpp/h`: Multi-class SAMME AdaBoost implementation
+- `DecisionTree.cpp/h`: Decision tree base classifier
+- `XA1DE.cpp/h`: Extended AODE variants
+- Experimental implementations of Bayesian network classifiers
+
+**Grid Search** (`src/grid/`):
+- `GridSearch.cpp/h`: Hyperparameter optimization
+- `GridExperiment.cpp/h`: Grid search experiment management
+- Uses MPI for parallel hyperparameter evaluation
+
+**Results & Reporting** (`src/results/`, `src/reports/`):
+- JSON-based result storage with schema validation
+- Excel export capabilities via libxlsxwriter
+- Console and paginated result display
+
+### Model Registration System
+
+The framework uses a factory pattern with automatic registration:
+- All classifiers inherit from `bayesnet::BaseClassifier`
+- Registration happens in `src/main/modelRegister.h`
+- Factory creates instances by string name via `Models::create()`
+
+## Configuration
+
+**Environment Configuration** (`.env` file):
+- `experiment`: Experiment name/type
+- `n_folds`: Cross-validation folds (default: 5)
+- `seeds`: Random seeds for reproducibility
+- `model`: Default classifier name
+- `score`: Primary evaluation metric
+- `platform`: System identifier for results
+
+**Grid Search Configuration**:
+- `grid_<model_name>_input.json`: Hyperparameter search space
+- `grid_<model_name>_output.json`: Search results
+
+## Data Format
+
+**Dataset Requirements**:
+- ARFF format files in `datasets/` directory
+- `all.txt` file listing datasets: `<name>,<class_name>,<real_features>`
+- Supports both discrete and continuous features
+- Automatic discretization available via MDLP
+
+**Experimental Data**:
+- Results stored in JSON format with versioned schemas
+- Test data in `tests/data/` for unit testing
+- Sample datasets: iris, diabetes, ecoli, glass, etc.
+
+## Development Workflow
+
+1. **Setup**: Run `make init` to install dependencies via Conan
+2. **Development**: Use `make debug` for development builds with testing
+3. **Testing**: Run `make test` after changes
+4. **Release**: Use `make release` for optimized builds
+5. **Experiments**: Use `.env` configuration and run `b_main` with appropriate flags
+
+## Key Features
+
+- **Multi-threaded**: Uses MPI for parallel grid search and experiments
+- **Cross-platform**: Supports Linux and macOS via vcpkg
+- **Extensible**: Easy classifier registration and integration
+- **Research-focused**: Designed for machine learning experimentation
+- **Visualization**: DOT graph generation for decision trees and networks

CMakeLists.txt

@@ -7,81 +7,68 @@ project(Platform
   LANGUAGES CXX
 )
 
-find_package(Torch REQUIRED)
-
-if (POLICY CMP0135)
-  cmake_policy(SET CMP0135 NEW)
-endif ()
-
 # Global CMake variables
 # ----------------------
-set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD 20)
 set(CMAKE_CXX_STANDARD_REQUIRED            ON)
 set(CMAKE_CXX_EXTENSIONS                  OFF)
 set(CMAKE_EXPORT_COMPILE_COMMANDS          ON)
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${TORCH_CXX_FLAGS}")
 SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pthread")
-set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast")
-set(CMAKE_CXX_FLAGS_DEBUG " ${CMAKE_CXX_FLAGS} -fprofile-arcs -ftest-coverage -O0 -g")
+set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -O3")
 
 # Options
 # -------
-option(ENABLE_CLANG_TIDY "Enable to add clang tidy."              OFF)
 option(ENABLE_TESTING "Unit testing build"                        OFF)
 option(CODE_COVERAGE "Collect coverage from test library"         OFF)
 
+# CMakes modules
+# --------------
+set(CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake/modules ${CMAKE_MODULE_PATH})
+
 # MPI
 find_package(MPI REQUIRED)
 message("MPI_CXX_LIBRARIES=${MPI_CXX_LIBRARIES}")
 message("MPI_CXX_INCLUDE_DIRS=${MPI_CXX_INCLUDE_DIRS}")
 
 # Boost Library
+cmake_policy(SET CMP0135 NEW)
+cmake_policy(SET CMP0167 NEW) # For FindBoost
 set(Boost_USE_STATIC_LIBS OFF) 
 set(Boost_USE_MULTITHREADED ON)  
 set(Boost_USE_STATIC_RUNTIME OFF) 
+
 find_package(Boost 1.66.0 REQUIRED COMPONENTS python3 numpy3) 
+
+# # Python
+find_package(Python3 REQUIRED COMPONENTS Development)
+
+# # Boost Python
+# find_package(boost_python${Python3_VERSION_MAJOR}${Python3_VERSION_MINOR} CONFIG REQUIRED COMPONENTS python${Python3_VERSION_MAJOR}${Python3_VERSION_MINOR})
+# # target_link_libraries(MyTarget PRIVATE Boost::python${Python3_VERSION_MAJOR}${Python3_VERSION_MINOR})
+
+
 if(Boost_FOUND)
     message("Boost_INCLUDE_DIRS=${Boost_INCLUDE_DIRS}")
+    message("Boost_LIBRARIES=${Boost_LIBRARIES}")
+    message("Boost_VERSION=${Boost_VERSION}")
     include_directories(${Boost_INCLUDE_DIRS}) 
 endif()
 
-# Python
-find_package(Python3 3.11 COMPONENTS Interpreter Development REQUIRED)
-message("Python3_LIBRARIES=${Python3_LIBRARIES}")
 
-# CMakes modules
-# --------------
-set(CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake/modules ${CMAKE_MODULE_PATH})
-include(AddGitSubmodule)
-
-if (CODE_COVERAGE)
-    enable_testing()
-    include(CodeCoverage)
-    MESSAGE("Code coverage enabled")
-    SET(GCC_COVERAGE_LINK_FLAGS " ${GCC_COVERAGE_LINK_FLAGS} -lgcov --coverage")
-endif (CODE_COVERAGE)
-
-if (ENABLE_CLANG_TIDY)
-  include(StaticAnalyzers) # clang-tidy
-endif (ENABLE_CLANG_TIDY)
 
 # External libraries - dependencies of Platform
 # ---------------------------------------------
-add_git_submodule("lib/argparse")
-add_git_submodule("lib/mdlp")
-
-find_library(XLSXWRITER_LIB NAMES libxlsxwriter.dylib libxlsxwriter.so PATHS ${Platform_SOURCE_DIR}/lib/libxlsxwriter/lib)
-message("XLSXWRITER_LIB=${XLSXWRITER_LIB}")
-
-find_library(PyClassifiers NAMES libPyClassifiers PyClassifiers libPyClassifiers.a PATHS ${Platform_SOURCE_DIR}/../lib/lib REQUIRED)
-find_path(PyClassifiers_INCLUDE_DIRS REQUIRED NAMES pyclassifiers PATHS ${Platform_SOURCE_DIR}/../lib/include)
-find_library(BayesNet NAMES libBayesNet BayesNet libBayesNet.a PATHS ${Platform_SOURCE_DIR}/../lib/lib REQUIRED)
-find_path(Bayesnet_INCLUDE_DIRS REQUIRED NAMES bayesnet PATHS ${Platform_SOURCE_DIR}/../lib/include)
-
-message(STATUS "PyClassifiers=${PyClassifiers}")
-message(STATUS "PyClassifiers_INCLUDE_DIRS=${PyClassifiers_INCLUDE_DIRS}")
-message(STATUS "BayesNet=${BayesNet}")
-message(STATUS "Bayesnet_INCLUDE_DIRS=${Bayesnet_INCLUDE_DIRS}")
+find_package(nlohmann_json CONFIG REQUIRED)
+find_package(argparse CONFIG REQUIRED)
+find_package(Torch CONFIG REQUIRED)
+find_package(arff-files CONFIG REQUIRED)
+find_package(fimdlp CONFIG REQUIRED)
+find_package(folding CONFIG REQUIRED)
+find_package(bayesnet CONFIG REQUIRED)
+find_package(pyclassifiers CONFIG REQUIRED)
+find_package(libxlsxwriter CONFIG REQUIRED)
+find_package(Boost REQUIRED COMPONENTS python)
 
 # Subdirectories
 # --------------
@@ -97,9 +84,15 @@ file(GLOB Platform_SOURCES CONFIGURE_DEPENDS ${Platform_SOURCE_DIR}/src/*.cpp)
 # -------
 if (ENABLE_TESTING)
   MESSAGE("Testing enabled")
-  if (NOT TARGET Catch2::Catch2)
-    add_git_submodule("lib/catch2")
-  endif (NOT TARGET Catch2::Catch2)
+  set(CMAKE_CXX_FLAGS_DEBUG " ${CMAKE_CXX_FLAGS} -fprofile-arcs -ftest-coverage -O0 -g")
+  enable_testing()
+  find_package(Catch2 CONFIG REQUIRED)
+  set(CODE_COVERAGE ON)
   include(CTest)
   add_subdirectory(tests)
 endif (ENABLE_TESTING)
+if (CODE_COVERAGE)
+    MESSAGE("Code coverage enabled")
+    include(CodeCoverage)
+    SET(GCC_COVERAGE_LINK_FLAGS " ${GCC_COVERAGE_LINK_FLAGS} -lgcov --coverage")
+endif (CODE_COVERAGE)

Makefile

@@ -1,11 +1,18 @@
 SHELL := /bin/bash
 .DEFAULT_GOAL := help
-.PHONY: coverage setup help build test clean debug release submodules buildr buildd install dependency testp testb clang-uml
+.PHONY: init clean coverage setup help build test clean debug release buildr buildd install dependency testp testb clang-uml example
 
-f_release = build_release
-f_debug = build_debug
-app_targets = b_best b_list b_main b_manage b_grid
+f_release = build_Release
+f_debug = build_Debug
+app_targets = b_best b_list b_main b_manage b_grid b_results
 test_targets = unit_tests_platform
+# Set the number of parallel jobs to the number of available processors minus 7
+CPUS := $(shell getconf _NPROCESSORS_ONLN 2>/dev/null \
+                 || nproc --all 2>/dev/null \
+                 || sysctl -n hw.ncpu)
+
+# --- Your desired job count: CPUs – 7, but never less than 1 --------------
+JOBS := $(shell n=$(CPUS); [ $${n} -gt 7 ] && echo $$((n-7)) || echo 1)
 
 define ClearTests
 	@for t in $(test_targets); do \
@@ -20,14 +27,43 @@ define ClearTests
 	fi ; 
 endef
 
+define build_target
+	@echo ">>> Building the project for $(1)..."
+	@if [ -d $(2) ]; then rm -fr $(2); fi
+	@conan install . --build=missing -of $(2) -s build_type=$(1)
+	@cmake -S . -B $(2) -DCMAKE_TOOLCHAIN_FILE=$(2)/build/$(1)/generators/conan_toolchain.cmake -DCMAKE_BUILD_TYPE=$(1) -D$(3)
+	@echo ">>> Will build using $(JOBS) parallel jobs"
+	echo ">>> Done"
+endef
 
-sub-init: ## Initialize submodules
-	@git submodule update --init --recursive
+define compile_target
+	@echo ">>> Compiling for $(1)..."
+		if [ "$(3)" != "" ]; then \
+		target="-t$(3)"; \
+	else \
+		target=""; \
+	fi
+	@cmake --build $(2) --config $(1) --parallel $(JOBS) $(target)
+	@echo ">>> Done"
+endef
 
-sub-update: ## Initialize submodules
-	@git submodule update --remote --merge
-	@git submodule foreach git pull origin master
+init: ## Initialize the project installing dependencies
+	@echo ">>> Installing dependencies with Conan"
+	@conan install . --output-folder=build --build=missing -s build_type=Release
+	@conan install . --output-folder=build_debug --build=missing -s build_type=Debug
+	@echo ">>> Done"
 
+clean: ## Clean the project
+	@echo ">>> Cleaning the project..."
+	@if test -f CMakeCache.txt ; then echo "- Deleting CMakeCache.txt"; rm -f CMakeCache.txt; fi
+	@for folder in $(f_release) $(f_debug) build build_debug install_test ; do \
+	if test -d "$$folder" ; then \
+		echo "- Deleting $$folder folder" ; \
+		rm -rf "$$folder"; \
+	fi; \
+	done
+	$(call ClearTests)
+	@echo ">>> Done";
 setup: ## Install dependencies for tests and coverage
 	@if [ "$(shell uname)" = "Darwin" ]; then \
 		brew install gcovr; \
@@ -46,7 +82,9 @@ install: ## Copy binary files to bin folder
 	@echo "*******************************************"
 	@for item in $(app_targets); do \
 		echo ">>> Copying $$item" ; \
-		cp $(f_release)/src/$$item $(dest) ; \
+		cp $(f_release)/src/$$item $(dest) || { \
+            echo "*** Error copying $$item" ; \
+        } ; \
 	done
 
 dependency: ## Create a dependency graph diagram of the project (build/dependency.png)
@@ -55,38 +93,27 @@ dependency: ## Create a dependency graph diagram of the project (build/dependenc
 	cd $(f_debug) && cmake .. --graphviz=dependency.dot && dot -Tpng dependency.dot -o dependency.png
 
 buildd: ## Build the debug targets
-	cmake --build $(f_debug) -t $(app_targets) PlatformSample --parallel
+	@$(call compile_target,"Debug","$(f_debug)")
 
 buildr: ## Build the release targets
-	cmake --build $(f_release) -t $(app_targets) --parallel
-
-clean: ## Clean the tests info
-	@echo ">>> Cleaning Debug Platform tests...";
-	$(call ClearTests)
-	@echo ">>> Done";
+	@$(call compile_target,"Release","$(f_release)")
 
 clang-uml: ## Create uml class and sequence diagrams
 	clang-uml -p --add-compile-flag -I /usr/lib/gcc/x86_64-redhat-linux/8/include/
 
-debug: ## Build a debug version of the project
-	@echo ">>> Building Debug Platform...";
-	@if [ -d ./$(f_debug) ]; then rm -rf ./$(f_debug); fi
-	@mkdir $(f_debug); 
-	@cmake -S . -B $(f_debug) -D CMAKE_BUILD_TYPE=Debug -D ENABLE_TESTING=ON -D CODE_COVERAGE=ON
-	@echo ">>> Done";
+debug: ## Build a debug version of the project with Conan
+	@$(call build_target,"Debug","$(f_debug)", "ENABLE_TESTING=ON")
+
+release: ## Build a Release version of the project with Conan
+	@$(call build_target,"Release","$(f_release)", "ENABLE_TESTING=OFF")
 
-release: ## Build a Release version of the project
-	@echo ">>> Building Release Platform...";
-	@if [ -d ./$(f_release) ]; then rm -rf ./$(f_release); fi
-	@mkdir $(f_release); 
-	@cmake -S . -B $(f_release) -D CMAKE_BUILD_TYPE=Release
-	@echo ">>> Done";	
 
 opt = ""
 test: ## Run tests (opt="-s") to verbose output the tests, (opt="-c='Test Maximum Spanning Tree'") to run only that section
 	@echo ">>> Running Platform tests...";
 	@$(MAKE) clean
-	@cmake --build $(f_debug) -t $(test_targets) --parallel
+	@$(MAKE) debug
+	@$(call "Compile_target", "Debug", "$(f_debug)", $(test_targets))
 	@for t in $(test_targets); do \
 		if [ -f $(f_debug)/tests/$$t ]; then \
 			cd $(f_debug)/tests ; \
@@ -98,8 +125,8 @@ test: ## Run tests (opt="-s") to verbose output the tests, (opt="-c='Test Maximu
 fname = iris
 example: ## Build sample
 	@echo ">>> Building Sample...";
-	@cmake --build build_debug -t sample
-	build_debug/sample/PlatformSample --model BoostAODE --dataset $(fname) --discretize --stratified
+	@cmake --build $(f_release) -t sample
+	$(f_release)/sample/PlatformSample --model BoostAODE --dataset $(fname) --discretize --stratified
 	@echo ">>> Done";
 
 

README.md

@@ -2,7 +2,8 @@
 
 ![C++](https://img.shields.io/badge/c++-%2300599C.svg?style=flat&logo=c%2B%2B&logoColor=white)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](<https://opensource.org/licenses/MIT>)
-![Gitea Last Commit](https://img.shields.io/gitea/last-commit/rmontanana/platform?gitea_url=https://gitea.rmontanana.es:3000&logo=gitea)
+[![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/rmontanana/Platform)
+![Gitea Last Commit](https://img.shields.io/gitea/last-commit/rmontanana/platform?gitea_url=https://gitea.rmontanana.es&logo=gitea)
 
 Platform to run Bayesian Networks and Machine Learning Classifiers experiments.
 
@@ -40,7 +41,7 @@ export MPI_HOME="/usr/lib64/openmpi"
 In Mac OS X, install mpich with brew and if cmake doesn't find it, edit mpicxx wrapper to remove the ",-commons,use_dylibs" from final_ldflags
 
 ```bash
-vi /opt/homebrew/bin/mpicx
+vi /opt/homebrew/bin/mpicxx
 ```
 
 ### boost library

cmake/modules/CodeCoverage.cmake

@@ -137,7 +137,7 @@
 
 include(CMakeParseArguments)
 
-option(CODE_COVERAGE_VERBOSE "Verbose information" FALSE)
+option(CODE_COVERAGE_VERBOSE "Verbose information" TRUE)
 
 # Check prereqs
 find_program( GCOV_PATH gcov )
@@ -160,7 +160,11 @@ foreach(LANG ${LANGUAGES})
     endif()
   elseif(NOT "${CMAKE_${LANG}_COMPILER_ID}" MATCHES "GNU"
          AND NOT "${CMAKE_${LANG}_COMPILER_ID}" MATCHES "(LLVM)?[Ff]lang")
-    message(FATAL_ERROR "Compiler is not GNU or Flang! Aborting...")
+         if ("${LANG}" MATCHES "CUDA")
+             message(STATUS "Ignoring CUDA")
+        else()
+            message(FATAL_ERROR "Compiler is not GNU or Flang! Aborting...")
+        endif()
   endif()
 endforeach()
 

conanfile.py

@@ -0,0 +1,42 @@
+from conan import ConanFile
+from conan.tools.cmake import CMakeToolchain, CMakeDeps, cmake_layout
+
+
+class PlatformConan(ConanFile):
+    name = "platform"
+    version = "1.1.0"
+    
+    # Binary configuration
+    settings = "os", "compiler", "build_type", "arch"
+    
+    # Sources are located in the same place as this recipe, copy them to the recipe
+    exports_sources = "CMakeLists.txt", "src/*", "tests/*", "config/*", "cmake/*"
+    
+    def requirements(conanself):
+        # Core dependencies from vcpkg.json
+        self.requires("argparse/3.2")
+        self.requires("libtorch/2.7.1")
+        self.requires("nlohmann_json/3.11.3")
+        self.requires("folding/1.1.2")
+        self.requires("fimdlp/2.1.1")
+        self.requires("arff-files/1.2.1")
+        self.requires("bayesnet/1.2.1")
+        self.requires("pyclassifiers/1.0.3")
+        self.requires("libxlsxwriter/1.2.2")
+        
+    def build_requirements(self):
+        self.tool_requires("cmake/[>=3.30]")
+        self.test_requires("catch2/3.8.1")
+    
+    def layout(self):
+        cmake_layout(self)
+    
+    def generate(self):
+        deps = CMakeDeps(self)
+        deps.generate()
+        tc = CMakeToolchain(self)
+        tc.generate()
+    
+    def configure(self):
+        # C++20 requirement
+        self.settings.compiler.cppstd = "20"

gitmodules

@@ -1,23 +0,0 @@
-[submodule "lib/catch2"]
-	path = lib/catch2
-        main  = v2.x
-	update = merge
-	url = https://github.com/catchorg/Catch2.git
-[submodule "lib/argparse"]
-	path = lib/argparse
-	url = https://github.com/p-ranav/argparse
-	master = master
-	update = merge
-[submodule "lib/json"]
-	path = lib/json
-	url = https://github.com/nlohmann/json.git
-        master = master
-	update = merge
-[submodule "lib/libxlsxwriter"]
-	path = lib/libxlsxwriter
-	url = https://github.com/jmcnamara/libxlsxwriter.git
-        main = main
-	update = merge
-[submodule "lib/folding"]
-	path = lib/folding
-	url = https://github.com/rmontanana/Folding

remove_submodules.sh

@@ -0,0 +1,14 @@
+git config --file .gitmodules --get-regexp path | awk '{ print $2 }' | while read line; do
+    echo "Removing $line"
+    # Deinit the submodule
+    git submodule deinit -f "$line"
+    
+    # Remove the submodule from the working tree
+    git rm -f "$line"
+    
+    # Remove the submodule from .git/modules
+    rm -rf ".git/modules/$line"
+done
+
+# Remove the .gitmodules file
+git rm -f .gitmodules

sample/CMakeLists.txt

@@ -1,15 +1,11 @@
 include_directories(
+    ${TORCH_INCLUDE_DIRS}
     ${Platform_SOURCE_DIR}/src/common
     ${Platform_SOURCE_DIR}/src/main
     ${Python3_INCLUDE_DIRS}
-    ${Platform_SOURCE_DIR}/lib/Files
-    ${Platform_SOURCE_DIR}/lib/mdlp/src
-    ${Platform_SOURCE_DIR}/lib/argparse/include
-    ${Platform_SOURCE_DIR}/lib/folding
-    ${Platform_SOURCE_DIR}/lib/json/include
     ${CMAKE_BINARY_DIR}/configured_files/include
     ${PyClassifiers_INCLUDE_DIRS}
-    ${Bayesnet_INCLUDE_DIRS}
+    ${bayesnet_INCLUDE_DIRS}
 )
 add_executable(PlatformSample sample.cpp ${Platform_SOURCE_DIR}/src/main/Models.cpp) 
-target_link_libraries(PlatformSample "${PyClassifiers}" "${BayesNet}" fimdlp ${Python3_LIBRARIES} "${TORCH_LIBRARIES}" ${LIBTORCH_PYTHON} Boost::python Boost::numpy)
+target_link_libraries(PlatformSample "${PyClassifiers}" "${BayesNet}" fimdlp ${Python3_LIBRARIES} "${TORCH_LIBRARIES}" ${LIBTORCH_PYTHON} ${Boost_LIBRARIES})

sample/sample.cpp

@@ -9,6 +9,7 @@
 #include <fimdlp/CPPFImdlp.h>
 #include <folding.hpp>
 #include <bayesnet/utils/BayesMetrics.h>
+#include <bayesnet/classifiers/SPODE.h>
 #include "Models.h"
 #include "modelRegister.h"
 #include "config_platform.h"
@@ -160,82 +161,119 @@ int main(int argc, char** argv)
             states[feature] = std::vector<int>(maxes[feature]);
         }
         states[className] = std::vector<int>(maxes[className]);
-        auto clf = platform::Models::instance()->create(model_name);
+        // Output the states
+        std::cout << std::string(80, '-') << std::endl;
+        std::cout << "States" << std::endl;
+        for (auto feature : features) {
+            std::cout << feature << ": " << states[feature].size() << std::endl;
+        }
+        std::cout << std::string(80, '-') << std::endl;
+        //auto clf = platform::Models::instance()->create("SPODE");
+        auto clf = bayesnet::SPODE(2);
+
         bayesnet::Smoothing_t smoothing = bayesnet::Smoothing_t::ORIGINAL;
-        clf->fit(Xd, y, features, className, states, smoothing);
+        clf.fit(Xd, y, features, className, states, smoothing);
         if (dump_cpt) {
             std::cout << "--- CPT Tables ---" << std::endl;
-            clf->dump_cpt();
+            std::cout << clf.dump_cpt();
         }
-        auto lines = clf->show();
+        std::cout << "--- Datos predicción ---" << std::endl;
+        std::cout << "Orden de variables: " << std::endl;
+        for (auto feature : features) {
+            std::cout << feature << ", ";
+        }
+        std::cout << std::endl;
+        std::cout << "X[0]: ";
+        for (int i = 0; i < Xd.size(); ++i) {
+            std::cout << Xd[i][0] << ", ";
+        }
+        std::cout << std::endl;
+        std::cout << std::string(80, '-') << std::endl;
+
+        auto lines = clf.show();
         for (auto line : lines) {
             std::cout << line << std::endl;
         }
         std::cout << "--- Topological Order ---" << std::endl;
-        auto order = clf->topological_order();
+        auto order = clf.topological_order();
         for (auto name : order) {
             std::cout << name << ", ";
         }
-        std::cout << "end." << std::endl;
-        auto score = clf->score(Xd, y);
-        std::cout << "Score: " << score << std::endl;
-        auto graph = clf->graph();
-        auto dot_file = model_name + "_" + file_name;
-        ofstream file(dot_file + ".dot");
-        file << graph;
-        file.close();
-        std::cout << "Graph saved in " << model_name << "_" << file_name << ".dot" << std::endl;
-        std::cout << "dot -Tpng -o " + dot_file + ".png " + dot_file + ".dot " << std::endl;
-        std::string stratified_string = stratified ? " Stratified" : "";
-        std::cout << nFolds << " Folds" << stratified_string << " Cross validation" << std::endl;
-        std::cout << "==========================================" << std::endl;
-        torch::Tensor Xt = torch::zeros({ static_cast<int>(Xd.size()), static_cast<int>(Xd[0].size()) }, torch::kInt32);
-        torch::Tensor yt = torch::tensor(y, torch::kInt32);
-        for (int i = 0; i < features.size(); ++i) {
-            Xt.index_put_({ i, "..." }, torch::tensor(Xd[i], torch::kInt32));
-        }
-        float total_score = 0, total_score_train = 0, score_train, score_test;
-        folding::Fold* fold;
-        double nodes = 0.0;
-        if (stratified)
-            fold = new folding::StratifiedKFold(nFolds, y, seed);
-        else
-            fold = new folding::KFold(nFolds, y.size(), seed);
-        for (auto i = 0; i < nFolds; ++i) {
-            auto [train, test] = fold->getFold(i);
-            std::cout << "Fold: " << i + 1 << std::endl;
-            if (tensors) {
-                auto ttrain = torch::tensor(train, torch::kInt64);
-                auto ttest = torch::tensor(test, torch::kInt64);
-                torch::Tensor Xtraint = torch::index_select(Xt, 1, ttrain);
-                torch::Tensor ytraint = yt.index({ ttrain });
-                torch::Tensor Xtestt = torch::index_select(Xt, 1, ttest);
-                torch::Tensor ytestt = yt.index({ ttest });
-                clf->fit(Xtraint, ytraint, features, className, states, smoothing);
-                auto temp = clf->predict(Xtraint);
-                score_train = clf->score(Xtraint, ytraint);
-                score_test = clf->score(Xtestt, ytestt);
-            } else {
-                auto [Xtrain, ytrain] = extract_indices(train, Xd, y);
-                auto [Xtest, ytest] = extract_indices(test, Xd, y);
-                clf->fit(Xtrain, ytrain, features, className, states, smoothing);
-                std::cout << "Nodes: " << clf->getNumberOfNodes() << std::endl;
-                nodes += clf->getNumberOfNodes();
-                score_train = clf->score(Xtrain, ytrain);
-                score_test = clf->score(Xtest, ytest);
+        auto predict_proba = clf.predict_proba(Xd);
+        std::cout << "Instances predict_proba: ";
+        for (int i = 0; i < predict_proba.size(); i++) {
+            std::cout << "Instance " << i << ": ";
+            for (int j = 0; j < 4; ++j) {
+                std::cout << Xd[j][i] << ", ";
             }
-            if (dump_cpt) {
-                std::cout << "--- CPT Tables ---" << std::endl;
-                clf->dump_cpt();
+            std::cout << ": ";
+            for (auto score : predict_proba[i]) {
+                std::cout << score << ", ";
             }
-            total_score_train += score_train;
-            total_score += score_test;
-            std::cout << "Score Train: " << score_train << std::endl;
-            std::cout << "Score Test : " << score_test << std::endl;
-            std::cout << "-------------------------------------------------------------------------------" << std::endl;
+            std::cout << std::endl;
         }
-        std::cout << "Nodes: " << nodes / nFolds << std::endl;
-        std::cout << "**********************************************************************************" << std::endl;
-        std::cout << "Average Score Train: " << total_score_train / nFolds << std::endl;
-        std::cout << "Average Score Test : " << total_score / nFolds << std::endl;return 0;
+        // std::cout << std::endl;
+        // std::cout << "end." << std::endl;
+        // auto score = clf->score(Xd, y);
+        // std::cout << "Score: " << score << std::endl;
+        // auto graph = clf->graph();
+        // auto dot_file = model_name + "_" + file_name;
+        // ofstream file(dot_file + ".dot");
+        // file << graph;
+        // file.close();
+        // std::cout << "Graph saved in " << model_name << "_" << file_name << ".dot" << std::endl;
+        // std::cout << "dot -Tpng -o " + dot_file + ".png " + dot_file + ".dot " << std::endl;
+        // std::string stratified_string = stratified ? " Stratified" : "";
+        // std::cout << nFolds << " Folds" << stratified_string << " Cross validation" << std::endl;
+        // std::cout << "==========================================" << std::endl;
+        // torch::Tensor Xt = torch::zeros({ static_cast<int>(Xd.size()), static_cast<int>(Xd[0].size()) }, torch::kInt32);
+        // torch::Tensor yt = torch::tensor(y, torch::kInt32);
+        // for (int i = 0; i < features.size(); ++i) {
+        //     Xt.index_put_({ i, "..." }, torch::tensor(Xd[i], torch::kInt32));
+        // }
+        // float total_score = 0, total_score_train = 0, score_train, score_test;
+        // folding::Fold* fold;
+        // double nodes = 0.0;
+        // if (stratified)
+        //     fold = new folding::StratifiedKFold(nFolds, y, seed);
+        // else
+        //     fold = new folding::KFold(nFolds, y.size(), seed);
+        // for (auto i = 0; i < nFolds; ++i) {
+        //     auto [train, test] = fold->getFold(i);
+        //     std::cout << "Fold: " << i + 1 << std::endl;
+        //     if (tensors) {
+        //         auto ttrain = torch::tensor(train, torch::kInt64);
+        //         auto ttest = torch::tensor(test, torch::kInt64);
+        //         torch::Tensor Xtraint = torch::index_select(Xt, 1, ttrain);
+        //         torch::Tensor ytraint = yt.index({ ttrain });
+        //         torch::Tensor Xtestt = torch::index_select(Xt, 1, ttest);
+        //         torch::Tensor ytestt = yt.index({ ttest });
+        //         clf->fit(Xtraint, ytraint, features, className, states, smoothing);
+        //         auto temp = clf->predict(Xtraint);
+        //         score_train = clf->score(Xtraint, ytraint);
+        //         score_test = clf->score(Xtestt, ytestt);
+        //     } else {
+        //         auto [Xtrain, ytrain] = extract_indices(train, Xd, y);
+        //         auto [Xtest, ytest] = extract_indices(test, Xd, y);
+        //         clf->fit(Xtrain, ytrain, features, className, states, smoothing);
+        //         std::cout << "Nodes: " << clf->getNumberOfNodes() << std::endl;
+        //         nodes += clf->getNumberOfNodes();
+        //         score_train = clf->score(Xtrain, ytrain);
+        //         score_test = clf->score(Xtest, ytest);
+        //     }
+        //     // if (dump_cpt) {
+        //     //     std::cout << "--- CPT Tables ---" << std::endl;
+        //     //     std::cout << clf->dump_cpt();
+        //     // }
+        //     total_score_train += score_train;
+        //     total_score += score_test;
+        //     std::cout << "Score Train: " << score_train << std::endl;
+        //     std::cout << "Score Test : " << score_test << std::endl;
+        //     std::cout << "-------------------------------------------------------------------------------" << std::endl;
+        // }
+
+        // std::cout << "Nodes: " << nodes / nFolds << std::endl;
+        // std::cout << "**********************************************************************************" << std::endl;
+        // std::cout << "Average Score Train: " << total_score_train / nFolds << std::endl;
+        // std::cout << "Average Score Test : " << total_score / nFolds << std::endl;return 0;
 }

src/CMakeLists.txt

@@ -1,20 +1,8 @@
 include_directories(
-    ## Libs
-    ${Platform_SOURCE_DIR}/lib/Files
-    ${Platform_SOURCE_DIR}/lib/folding
-    ${Platform_SOURCE_DIR}/lib/mdlp/src
-    ${Platform_SOURCE_DIR}/lib/argparse/include
-    ${Platform_SOURCE_DIR}/lib/json/include
-    ${Platform_SOURCE_DIR}/lib/libxlsxwriter/include
     ${Python3_INCLUDE_DIRS}
     ${MPI_CXX_INCLUDE_DIRS}
-    ${TORCH_INCLUDE_DIRS}
     ${CMAKE_BINARY_DIR}/configured_files/include
-    ${PyClassifiers_INCLUDE_DIRS}
-    ${Bayesnet_INCLUDE_DIRS}
-    ## Platform
     ${Platform_SOURCE_DIR}/src
-    ${Platform_SOURCE_DIR}/results
 )
 
 # b_best
@@ -25,17 +13,27 @@ add_executable(
     main/Models.cpp main/Scores.cpp
     reports/ReportExcel.cpp reports/ReportBase.cpp reports/ExcelFile.cpp
     results/Result.cpp
+    experimental_clfs/XA1DE.cpp
+    experimental_clfs/ExpClf.cpp
+    experimental_clfs/DecisionTree.cpp
+    experimental_clfs/AdaBoost.cpp
 )
-target_link_libraries(b_best Boost::boost "${PyClassifiers}" "${BayesNet}" fimdlp ${Python3_LIBRARIES} "${TORCH_LIBRARIES}" ${LIBTORCH_PYTHON} Boost::python Boost::numpy "${XLSXWRITER_LIB}")
+target_link_libraries(b_best Boost::boost pyclassifiers::pyclassifiers bayesnet::bayesnet argparse::argparse fimdlp::fimdlp ${Python3_LIBRARIES} torch::torch Boost::python Boost::numpy libxlsxwriter::libxlsxwriter)
 
 # b_grid
-set(grid_sources GridSearch.cpp GridData.cpp)
+set(grid_sources GridSearch.cpp GridData.cpp GridExperiment.cpp GridBase.cpp )
 list(TRANSFORM grid_sources PREPEND grid/)
 add_executable(b_grid commands/b_grid.cpp ${grid_sources} 
     common/Datasets.cpp common/Dataset.cpp common/Discretization.cpp
-    main/HyperParameters.cpp main/Models.cpp 
+    main/HyperParameters.cpp main/Models.cpp main/Experiment.cpp main/Scores.cpp main/ArgumentsExperiment.cpp
+    reports/ReportConsole.cpp reports/ReportBase.cpp 
+    results/Result.cpp
+    experimental_clfs/XA1DE.cpp
+    experimental_clfs/ExpClf.cpp
+    experimental_clfs/DecisionTree.cpp
+    experimental_clfs/AdaBoost.cpp
 )
-target_link_libraries(b_grid ${MPI_CXX_LIBRARIES} "${PyClassifiers}" "${BayesNet}" fimdlp ${Python3_LIBRARIES} "${TORCH_LIBRARIES}" ${LIBTORCH_PYTHON} Boost::python Boost::numpy)
+target_link_libraries(b_grid ${MPI_CXX_LIBRARIES} pyclassifiers::pyclassifiers bayesnet::bayesnet argparse::argparse fimdlp::fimdlp ${Python3_LIBRARIES} torch::torch Boost::python Boost::numpy)
 
 # b_list
 add_executable(b_list commands/b_list.cpp
@@ -43,18 +41,27 @@ add_executable(b_list commands/b_list.cpp
     main/Models.cpp main/Scores.cpp
     reports/ReportExcel.cpp reports/ExcelFile.cpp reports/ReportBase.cpp reports/DatasetsExcel.cpp reports/DatasetsConsole.cpp reports/ReportsPaged.cpp
     results/Result.cpp results/ResultsDatasetExcel.cpp results/ResultsDataset.cpp results/ResultsDatasetConsole.cpp
+    experimental_clfs/XA1DE.cpp
+    experimental_clfs/ExpClf.cpp
+    experimental_clfs/DecisionTree.cpp
+    experimental_clfs/AdaBoost.cpp
 )
-target_link_libraries(b_list "${PyClassifiers}" "${BayesNet}" fimdlp ${Python3_LIBRARIES} "${TORCH_LIBRARIES}" ${LIBTORCH_PYTHON} Boost::python Boost::numpy "${XLSXWRITER_LIB}")
+target_link_libraries(b_list pyclassifiers::pyclassifiers bayesnet::bayesnet argparse::argparse fimdlp::fimdlp ${Python3_LIBRARIES} torch::torch Boost::python Boost::numpy libxlsxwriter::libxlsxwriter)
 
 # b_main
-set(main_sources Experiment.cpp Models.cpp HyperParameters.cpp Scores.cpp)
+set(main_sources Experiment.cpp Models.cpp HyperParameters.cpp Scores.cpp ArgumentsExperiment.cpp)
 list(TRANSFORM main_sources PREPEND main/)
 add_executable(b_main commands/b_main.cpp ${main_sources} 
     common/Datasets.cpp common/Dataset.cpp common/Discretization.cpp
     reports/ReportConsole.cpp reports/ReportBase.cpp 
     results/Result.cpp
+    experimental_clfs/XA1DE.cpp
+    experimental_clfs/ExpClf.cpp
+    experimental_clfs/ExpClf.cpp
+    experimental_clfs/DecisionTree.cpp
+    experimental_clfs/AdaBoost.cpp
 )
-target_link_libraries(b_main "${PyClassifiers}" "${BayesNet}" fimdlp ${Python3_LIBRARIES} "${TORCH_LIBRARIES}" ${LIBTORCH_PYTHON} Boost::python Boost::numpy)
+target_link_libraries(b_main PRIVATE nlohmann_json::nlohmann_json pyclassifiers::pyclassifiers bayesnet::bayesnet argparse::argparse fimdlp::fimdlp ${Python3_LIBRARIES} torch::torch Boost::python Boost::numpy)
 
 # b_manage
 set(manage_sources ManageScreen.cpp OptionsMenu.cpp ResultsManager.cpp)
@@ -66,4 +73,8 @@ add_executable(
     results/Result.cpp results/ResultsDataset.cpp results/ResultsDatasetConsole.cpp
     main/Scores.cpp
 )
-target_link_libraries(b_manage "${TORCH_LIBRARIES}" "${XLSXWRITER_LIB}" fimdlp "${BayesNet}")
+target_link_libraries(b_manage torch::torch libxlsxwriter::libxlsxwriter fimdlp::fimdlp bayesnet::bayesnet argparse::argparse)
+
+# b_results
+add_executable(b_results commands/b_results.cpp)
+target_link_libraries(b_results torch::torch libxlsxwriter::libxlsxwriter fimdlp::fimdlp bayesnet::bayesnet argparse::argparse)

src/best/BestResults.cpp

@@ -4,6 +4,7 @@
 #include <iostream>
 #include <sstream>
 #include <algorithm>
+#include <cctype>
 #include "common/Colors.h"
 #include "common/CLocale.h"
 #include "common/Paths.h"
@@ -123,15 +124,24 @@ namespace platform {
         }
         result = std::vector<std::string>(models.begin(), models.end());
         maxModelName = (*max_element(result.begin(), result.end(), [](const std::string& a, const std::string& b) { return a.size() < b.size(); })).size();
-        maxModelName = std::max(12, maxModelName);
+        maxModelName = std::max(minLength, maxModelName);
         return result;
     }
+    std::string toLower(std::string data)
+    {
+        std::transform(data.begin(), data.end(), data.begin(),
+            [](unsigned char c) { return std::tolower(c); });
+        return data;
+    }
     std::vector<std::string> BestResults::getDatasets(json table)
     {
         std::vector<std::string> datasets;
         for (const auto& dataset_ : table.items()) {
             datasets.push_back(dataset_.key());
         }
+        std::stable_sort(datasets.begin(), datasets.end(), [](const std::string& a, const std::string& b) {
+            return toLower(a) < toLower(b);
+            });
         maxDatasetName = (*max_element(datasets.begin(), datasets.end(), [](const std::string& a, const std::string& b) { return a.size() < b.size(); })).size();
         maxDatasetName = std::max(7, maxDatasetName);
         return datasets;
@@ -214,17 +224,17 @@ namespace platform {
         return table;
     }
 
-    void BestResults::printTableResults(std::vector<std::string> models, json table, bool tex)
+    void BestResults::printTableResults(std::vector<std::string> models, json table, bool tex, bool index)
     {
         std::stringstream oss;
         oss << Colors::GREEN() << "Best results for " << score << " as of " << table.at("dateTable").get<std::string>() << std::endl;
         std::cout << oss.str();
         std::cout << std::string(oss.str().size() - 8, '-') << std::endl;
         std::cout << Colors::GREEN() << " #  " << std::setw(maxDatasetName + 1) << std::left << std::string("Dataset");
-        auto bestResultsTex = BestResultsTex();
+        auto bestResultsTex = BestResultsTex(score);
         auto bestResultsMd = BestResultsMd();
         if (tex) {
-            bestResultsTex.results_header(models, table.at("dateTable").get<std::string>());
+            bestResultsTex.results_header(models, table.at("dateTable").get<std::string>(), index);
             bestResultsMd.results_header(models, table.at("dateTable").get<std::string>());
         }
         for (const auto& model : models) {
@@ -241,7 +251,7 @@ namespace platform {
         int nDatasets = table.begin().value().size();
         auto datasets = getDatasets(table.begin().value());
         if (tex) {
-            bestResultsTex.results_body(datasets, table);
+            bestResultsTex.results_body(datasets, table, index);
             bestResultsMd.results_body(datasets, table);
         }
         for (auto const& dataset_ : datasets) {
@@ -265,12 +275,14 @@ namespace platform {
             // Print the row with red colors on max values
             for (const auto& model : models) {
                 std::string efectiveColor = color;
-                double value;
+                double value, std;
                 try {
                     value = table[model].at(dataset_).at(0).get<double>();
+                    std = table[model].at(dataset_).at(3).get<double>();
                 }
                 catch (nlohmann::json_abi_v3_11_3::detail::out_of_range err) {
                     value = -1.0;
+                    std = -1.0;
                 }
                 if (value == maxValue) {
                     efectiveColor = Colors::RED();
@@ -279,7 +291,8 @@ namespace platform {
                     std::cout << Colors::YELLOW() << std::setw(maxModelName) << std::right << "N/A" << " ";
                 } else {
                     totals[model].push_back(value);
-                    std::cout << efectiveColor << std::setw(maxModelName) << std::setprecision(maxModelName - 2) << std::fixed << value << " ";
+                    std::cout << efectiveColor << std::setw(maxModelName - 6) << std::setprecision(maxModelName - 8) << std::fixed << value;
+                    std::cout << efectiveColor << "±" << std::setw(5) << std::setprecision(3) << std::fixed << std << " ";
                 }
             }
             std::cout << std::endl;
@@ -306,9 +319,9 @@ namespace platform {
         for (const auto& model : models) {
             std::string efectiveColor = model == best_model ? Colors::RED() : Colors::GREEN();
             double value = std::reduce(totals[model].begin(), totals[model].end()) / nDatasets;
-            double std_value = compute_std(totals[model], value);
-            std::cout << efectiveColor << std::right << std::setw(maxModelName) << std::setprecision(maxModelName - 4) << std::fixed << value << " ";
-
+            double std = compute_std(totals[model], value);
+            std::cout << efectiveColor << std::right << std::setw(maxModelName - 6) << std::setprecision(maxModelName - 8) << std::fixed << value;
+            std::cout << efectiveColor << "±" << std::setw(5) << std::setprecision(3) << std::fixed << std << " ";
         }
         std::cout << std::endl;
     }
@@ -320,25 +333,27 @@ namespace platform {
             // Build the table of results
             json table = buildTableResults(models);
             std::vector<std::string> datasets = getDatasets(table.begin().value());
-            BestResultsExcel excel_report(score, datasets);
+            BestResultsExcel excel_report(path, score, datasets);
             excel_report.reportSingle(model, path + Paths::bestResultsFile(score, model));
             messageOutputFile("Excel", excel_report.getFileName());
+            excelFileName = excel_report.getFileName();
         }
     }
-    void BestResults::reportAll(bool excel, bool tex)
+    void BestResults::reportAll(bool excel, bool tex, bool index)
     {
         auto models = getModels();
         // Build the table of results
         json table = buildTableResults(models);
         std::vector<std::string> datasets = getDatasets(table.begin().value());
         // Print the table of results
-        printTableResults(models, table, tex);
+        printTableResults(models, table, tex, index);
         // Compute the Friedman test
         std::map<std::string, std::map<std::string, float>> ranksModels;
         if (friedman) {
-            Statistics stats(models, datasets, table, significance);
+            Statistics stats(score, models, datasets, table, significance);
             auto result = stats.friedmanTest();
-            stats.postHocHolmTest(result, tex);
+            stats.postHocTest();
+            stats.postHocTestReport(result, tex);
             ranksModels = stats.getRanks();
         }
         if (tex) {
@@ -350,33 +365,21 @@ namespace platform {
             }
         }
         if (excel) {
-            BestResultsExcel excel(score, datasets);
+            BestResultsExcel excel(path, score, datasets);
             excel.reportAll(models, table, ranksModels, friedman, significance);
             if (friedman) {
-                int idx = -1;
-                double min = 2000;
-                // Find out the control model
-                auto totals = std::vector<double>(models.size(), 0.0);
-                for (const auto& dataset_ : datasets) {
-                    for (int i = 0; i < models.size(); ++i) {
-                        totals[i] += ranksModels[dataset_][models[i]];
-                    }
-                }
-                for (int i = 0; i < models.size(); ++i) {
-                    if (totals[i] < min) {
-                        min = totals[i];
-                        idx = i;
-                    }
-                }
+                Statistics stats(score, models, datasets, table, significance);
+                int idx = stats.getControlIdx();
                 model = models.at(idx);
                 excel.reportSingle(model, path + Paths::bestResultsFile(score, model));
             }
             messageOutputFile("Excel", excel.getFileName());
+            excelFileName = excel.getFileName();
         }
     }
     void BestResults::messageOutputFile(const std::string& title, const std::string& fileName)
     {
-        std::cout << Colors::YELLOW() << "** " << std::setw(5) << std::left << title
+        std::cout << Colors::YELLOW() << "** " << std::setw(8) << std::left << title
             << " file generated: " << fileName << Colors::RESET() << std::endl;
     }
 }

src/best/BestResults.h

@@ -13,15 +13,16 @@ namespace platform {
         }
         std::string build();
         void reportSingle(bool excel);
-        void reportAll(bool excel, bool tex);
+        void reportAll(bool excel, bool tex, bool index);
         void buildAll();
+        std::string getExcelFileName() const { return excelFileName; }
     private:
         std::vector<std::string> getModels();
         std::vector<std::string> getDatasets(json table);
         std::vector<std::string> loadResultFiles();
         void messageOutputFile(const std::string& title, const std::string& fileName);
         json buildTableResults(std::vector<std::string> models);
-        void printTableResults(std::vector<std::string> models, json table, bool tex);
+        void printTableResults(std::vector<std::string> models, json table, bool tex, bool index);
         json loadFile(const std::string& fileName);
         void listFile();
         std::string path;
@@ -32,6 +33,8 @@ namespace platform {
         double significance;
         int maxModelName = 0;
         int maxDatasetName = 0;
+        int minLength = 13; // Minimum length for scores
+        std::string excelFileName;
     };
 }
 #endif

src/best/BestResultsExcel.cpp

@@ -30,7 +30,7 @@ namespace platform {
         }
         return columnName;
     }
-    BestResultsExcel::BestResultsExcel(const std::string& score, const std::vector<std::string>& datasets) : score(score), datasets(datasets)
+    BestResultsExcel::BestResultsExcel(const std::string& path, const std::string& score, const std::vector<std::string>& datasets) : path(path), score(score), datasets(datasets)
     {
         file_name = Paths::bestResultsExcel(score);
         workbook = workbook_new(getFileName().c_str());
@@ -92,7 +92,7 @@ namespace platform {
             catch (const std::out_of_range& oor) {
                 auto tabName = "table_" + std::to_string(i);
                 auto worksheetNew = workbook_add_worksheet(workbook, tabName.c_str());
-                json data = loadResultData(Paths::results() + fileName);
+                json data = loadResultData(path + fileName);
                 auto report = ReportExcel(data, false, workbook, worksheetNew);
                 report.show();
                 hyperlink = "#table_" + std::to_string(i);
@@ -164,13 +164,15 @@ namespace platform {
         addConditionalFormat("max");
         footer(false);
         if (friedman) {
-            // Create Sheet with ranks
-            worksheet = workbook_add_worksheet(workbook, "Ranks");
-            formatColumns();
-            header(true);
-            body(true);
-            addConditionalFormat("min");
-            footer(true);
+            if (score == "accuracy") {
+                // Create Sheet with ranks
+                worksheet = workbook_add_worksheet(workbook, "Ranks");
+                formatColumns();
+                header(true);
+                body(true);
+                addConditionalFormat("min");
+                footer(true);
+            }
             // Create Sheet with Friedman Test
             doFriedman();
         }
@@ -241,11 +243,12 @@ namespace platform {
         }
         worksheet_merge_range(worksheet, 0, 0, 0, 7, "Friedman Test", styles["headerFirst"]);
         row = 2;
-        Statistics stats(models, datasets, table, significance, false);
+        Statistics stats(score, models, datasets, table, significance, false); // No output
         auto result = stats.friedmanTest();
-        stats.postHocHolmTest(result);
+        stats.postHocTest();
+        stats.postHocTestReport(result, false); // No tex output
         auto friedmanResult = stats.getFriedmanResult();
-        auto holmResult = stats.getHolmResult();
+        auto postHocResults = stats.getPostHocResults();
         worksheet_merge_range(worksheet, row, 0, row, 7, "Null hypothesis: H0 'There is no significant differences between all the classifiers.'", styles["headerSmall"]);
         row += 2;
         writeString(row, 1, "Friedman Q", "bodyHeader");
@@ -264,7 +267,7 @@ namespace platform {
         row += 2;
         worksheet_merge_range(worksheet, row, 0, row, 7, "Null hypothesis: H0 'There is no significant differences between the control model and the other models.'", styles["headerSmall"]);
         row += 2;
-        std::string controlModel = "Control Model: " + holmResult.model;
+        std::string controlModel = "Control Model: " + postHocResults.at(0).model;
         worksheet_merge_range(worksheet, row, 1, row, 7, controlModel.c_str(), styles["bodyHeader_odd"]);
         row++;
         writeString(row, 1, "Model", "bodyHeader");
@@ -276,7 +279,7 @@ namespace platform {
         writeString(row, 7, "Reject H0", "bodyHeader");
         row++;
         bool first = true;
-        for (const auto& item : holmResult.holmLines) {
+        for (const auto& item : postHocResults) {
             writeString(row, 1, item.model, "text");
             if (first) {
                 // Control model info

src/best/BestResultsExcel.h

@@ -10,7 +10,7 @@ namespace platform {
     using json = nlohmann::ordered_json;
     class BestResultsExcel : public ExcelFile {
     public:
-        BestResultsExcel(const std::string& score, const std::vector<std::string>& datasets);
+        BestResultsExcel(const std::string& path, const std::string& score, const std::vector<std::string>& datasets);
         ~BestResultsExcel();
         void reportAll(const std::vector<std::string>& models, const json& table, const std::map<std::string, std::map<std::string, float>>& ranks, bool friedman, double significance);
         void reportSingle(const std::string& model, const std::string& fileName);
@@ -22,6 +22,7 @@ namespace platform {
         void formatColumns();
         void doFriedman();
         void addConditionalFormat(std::string formula);
+        std::string path;
         std::string score;
         std::vector<std::string> models;
         std::vector<std::string> datasets;

src/best/BestResultsMd.cpp

@@ -75,7 +75,7 @@ namespace platform {
 
         handler.close();
     }
-    void BestResultsMd::holm_test(struct HolmResult& holmResult, const std::string& date)
+    void BestResultsMd::postHoc_test(std::vector<PostHocLine>& postHocResults, const std::string& kind, const std::string& date)
     {
         auto file_name = Paths::tex() + Paths::md_post_hoc();
         openMdFile(file_name);
@@ -84,13 +84,15 @@ namespace platform {
         handler << std::endl;
         handler << "  Post-hoc handler test" << std::endl;
         handler << "-->" << std::endl;
-        handler << "Post-hoc Holm test: H<sub>0</sub>: There is no significant differences between the control model and the other models." << std::endl << std::endl;
+        handler << "Post-hoc " << kind << " test: H<sub>0</sub>: There is no significant differences between the control model and the other models." << std::endl << std::endl;
         handler << "| classifier | pvalue | rank | win | tie | loss | H<sub>0</sub> |" << std::endl;
         handler << "| :-- | --: | --: | --:| --: | --: | :--: |" << std::endl;
-        for (auto const& line : holmResult.holmLines) {
+        bool first = true;
+        for (auto const& line : postHocResults) {
             auto textStatus = !line.reject ? "**" : " ";
-            if (line.model == holmResult.model) {
+            if (first) {
                 handler << "| " << line.model << " | - | " << std::fixed << std::setprecision(2) << line.rank << " | - | - | - |" << std::endl;
+                first = false;
             } else {
                 handler << "| " << line.model << " | " << textStatus << std::scientific << std::setprecision(4) << line.pvalue << textStatus << " |";
                 handler << std::fixed << std::setprecision(2) << line.rank << " | " << line.wtl.win << " | " << line.wtl.tie << " | " << line.wtl.loss << " |";

src/best/BestResultsMd.h

@@ -14,7 +14,7 @@ namespace platform {
         void results_header(const std::vector<std::string>& models, const std::string& date);
         void results_body(const std::vector<std::string>& datasets, json& table);
         void results_footer(const std::map<std::string, std::vector<double>>& totals, const std::string& best_model);
-        void holm_test(struct HolmResult& holmResult, const std::string& date);
+        void postHoc_test(std::vector<PostHocLine>& postHocResults, const std::string& kind, const std::string& date);
     private:
         void openMdFile(const std::string& name);
         std::ofstream handler;

src/best/BestResultsTex.cpp

@@ -12,7 +12,7 @@ namespace platform {
             exit(1);
         }
     }
-    void BestResultsTex::results_header(const std::vector<std::string>& models, const std::string& date)
+    void BestResultsTex::results_header(const std::vector<std::string>& models, const std::string& date, bool index)
     {
         this->models = models;
         auto file_name = Paths::tex() + Paths::tex_output();
@@ -27,9 +27,12 @@ namespace platform {
         handler << "\\tiny " << std::endl;
         handler << "\\renewcommand{\\arraystretch }{1.2} " << std::endl;
         handler << "\\renewcommand{\\tabcolsep }{0.07cm} " << std::endl;
-        handler << "\\caption{Accuracy results(mean $\\pm$ std) for all the algorithms and datasets} " << std::endl;
-        handler << "\\label{tab:results_accuracy}" << std::endl;
-        handler << "\\begin{tabular} {{r" << std::string(models.size(), 'c').c_str() << "}}" << std::endl;
+        auto umetric = score;
+        umetric[0] = toupper(umetric[0]);
+        handler << "\\caption{" << umetric << " results(mean $\\pm$ std) for all the algorithms and datasets} " << std::endl;
+        handler << "\\label{tab:results_" << score << "}" << std::endl;
+        std::string header_dataset_name = index ? "r" : "l";
+        handler << "\\begin{tabular} {{" << header_dataset_name << std::string(models.size(), 'c').c_str() << "}}" << std::endl;
         handler << "\\hline " << std::endl;
         handler << "" << std::endl;
         for (const auto& model : models) {
@@ -38,13 +41,12 @@ namespace platform {
         handler << "\\\\" << std::endl;
         handler << "\\hline" << std::endl;
     }
-    void BestResultsTex::results_body(const std::vector<std::string>& datasets, json& table)
+    void BestResultsTex::results_body(const std::vector<std::string>& datasets, json& table, bool index)
     {
         int i = 0;
         for (auto const& dataset : datasets) {
             // Find out max value for this dataset
             double max_value = 0;
-            // Find out the max value for this dataset
             for (const auto& model : models) {
                 double value;
                 try {
@@ -57,7 +59,10 @@ namespace platform {
                     max_value = value;
                 }
             }
-            handler << ++i << " ";
+            if (index)
+                handler << ++i << " ";
+            else
+                handler << dataset << " ";
             for (const auto& model : models) {
                 double value = table[model].at(dataset).at(0).get<double>();
                 double std_value = table[model].at(dataset).at(3).get<double>();
@@ -84,26 +89,28 @@ namespace platform {
         handler << "\\end{table}" << std::endl;
         handler.close();
     }
-    void BestResultsTex::holm_test(struct HolmResult& holmResult, const std::string& date)
+    void BestResultsTex::postHoc_test(std::vector<PostHocLine>& postHocResults, const std::string& kind, const std::string& date)
     {
         auto file_name = Paths::tex() + Paths::tex_post_hoc();
         openTexFile(file_name);
         handler << "%% This file has been generated by the platform program" << std::endl;
         handler << "%% Date: " << date.c_str() << std::endl;
         handler << "%%" << std::endl;
-        handler << "%% Post-hoc handler test" << std::endl;
+        handler << "%% Post-hoc " << kind << " test" << std::endl;
         handler << "%%" << std::endl;
         handler << "\\begin{table}[htbp]" << std::endl;
         handler << "\\centering" << std::endl;
-        handler << "\\caption{Results of the post-hoc test for the mean accuracy of the algorithms.}\\label{tab:tests}" << std::endl;
+        handler << "\\caption{Results of the post-hoc " << kind << " test for the mean " << score << " of the algorithms.}\\label{ tab:tests }" << std::endl;
         handler << "\\begin{tabular}{lrrrrr}" << std::endl;
         handler << "\\hline" << std::endl;
         handler << "classifier & pvalue & rank & win & tie & loss\\\\" << std::endl;
         handler << "\\hline" << std::endl;
-        for (auto const& line : holmResult.holmLines) {
+        bool first = true;
+        for (auto const& line : postHocResults) {
             auto textStatus = !line.reject ? "\\bf " : " ";
-            if (line.model == holmResult.model) {
+            if (first) {
                 handler << line.model << " & - & " << std::fixed << std::setprecision(2) << line.rank << " & - & - & - \\\\" << std::endl;
+                first = false;
             } else {
                 handler << line.model << " & " << textStatus << std::scientific << std::setprecision(4) << line.pvalue << " & ";
                 handler << std::fixed << std::setprecision(2) << line.rank << " & " << line.wtl.win << " & " << line.wtl.tie << " & " << line.wtl.loss << "\\\\" << std::endl;

src/best/BestResultsTex.h

@@ -9,13 +9,15 @@ namespace platform {
     using json = nlohmann::ordered_json;
     class BestResultsTex {
     public:
-        BestResultsTex() = default;
+        BestResultsTex(const std::string score, bool dataset_name = true) : score{ score }, dataset_name{ dataset_name } {};
         ~BestResultsTex() = default;
-        void results_header(const std::vector<std::string>& models, const std::string& date);
-        void results_body(const std::vector<std::string>& datasets, json& table);
+        void results_header(const std::vector<std::string>& models, const std::string& date, bool index);
+        void results_body(const std::vector<std::string>& datasets, json& table, bool index);
         void results_footer(const std::map<std::string, std::vector<double>>& totals, const std::string& best_model);
-        void holm_test(struct HolmResult& holmResult, const std::string& date);
+        void postHoc_test(std::vector<PostHocLine>& postHocResults, const std::string& kind, const std::string& date);
     private:
+        std::string score;
+        bool dataset_name;
         void openTexFile(const std::string& name);
         std::ofstream handler;
         std::vector<std::string> models;

src/best/Statistics.cpp

@@ -7,18 +7,25 @@
 #include "BestResultsTex.h"
 #include "BestResultsMd.h"
 #include "Statistics.h"
+#include "WilcoxonTest.hpp"
 
 
 namespace platform {
 
-    Statistics::Statistics(const std::vector<std::string>& models, const std::vector<std::string>& datasets, const json& data, double significance, bool output) :
-        models(models), datasets(datasets), data(data), significance(significance), output(output)
+    Statistics::Statistics(const std::string& score, const std::vector<std::string>& models, const std::vector<std::string>& datasets, const json& data, double significance, bool output) :
+        score(score), models(models), datasets(datasets), data(data), significance(significance), output(output)
     {
+        if (score == "accuracy") {
+            postHocType = "Holm";
+            hlen = 85;
+        } else {
+            postHocType = "Wilcoxon";
+            hlen = 88;
+        }
         nModels = models.size();
         nDatasets = datasets.size();
         auto temp = ConfigLocale();
     }
-
     void Statistics::fit()
     {
         if (nModels < 3 || nDatasets < 3) {
@@ -27,9 +34,11 @@ namespace platform {
             throw std::runtime_error("Can't make the Friedman test with less than 3 models and/or less than 3 datasets.");
         }
         ranksModels.clear();
-        computeRanks();
+        computeRanks(); // compute greaterAverage and ranks
         // Set the control model as the one with the lowest average rank
-        controlIdx = distance(ranks.begin(), min_element(ranks.begin(), ranks.end(), [](const auto& l, const auto& r) { return l.second < r.second; }));
+        controlIdx = score == "accuracy" ?
+            distance(ranks.begin(), min_element(ranks.begin(), ranks.end(), [](const auto& l, const auto& r) { return l.second < r.second; }))
+            : greaterAverage; // The model with the greater average score
         computeWTL();
         maxModelName = (*std::max_element(models.begin(), models.end(), [](const std::string& a, const std::string& b) { return a.size() < b.size(); })).size();
         maxDatasetName = (*std::max_element(datasets.begin(), datasets.end(), [](const std::string& a, const std::string& b) { return a.size() < b.size(); })).size();
@@ -66,11 +75,16 @@ namespace platform {
     void Statistics::computeRanks()
     {
         std::map<std::string, float> ranksLine;
+        std::map<std::string, float> averages;
+        for (const auto& model : models) {
+            averages[model] = 0;
+        }
         for (const auto& dataset : datasets) {
             std::vector<std::pair<std::string, double>> ranksOrder;
             for (const auto& model : models) {
                 double value = data[model].at(dataset).at(0).get<double>();
                 ranksOrder.push_back({ model, value });
+                averages[model] += value;
             }
             // Assign the ranks
             ranksLine = assignRanks(ranksOrder);
@@ -88,10 +102,17 @@ namespace platform {
         for (const auto& rank : ranks) {
             ranks[rank.first] /= nDatasets;
         }
+        // Average the scores
+        for (const auto& average : averages) {
+            averages[average.first] /= nDatasets;
+        }
+        // Get the model with the greater average score
+        greaterAverage = distance(averages.begin(), max_element(averages.begin(), averages.end(), [](const auto& l, const auto& r) { return l.second < r.second; }));
     }
     void Statistics::computeWTL()
     {
-        // Compute the WTL matrix
+        const double practical_threshold = 0.0005;
+        // Compute the WTL matrix (Win Tie Loss)
         for (int i = 0; i < nModels; ++i) {
             wtl[i] = { 0, 0, 0 };
         }
@@ -104,23 +125,85 @@ namespace platform {
                     continue;
                 }
                 double value = data[models[i]].at(item.key()).at(0).get<double>();
-                if (value < controlValue) {
-                    wtl[i].win++;
-                } else if (value == controlValue) {
+                double diff = controlValue - value; // control − comparison
+                if (std::fabs(diff) <= practical_threshold) {
                     wtl[i].tie++;
+                } else if (diff < 0) {
+                    wtl[i].win++;
                 } else {
                     wtl[i].loss++;
                 }
             }
         }
     }
-
-    void Statistics::postHocHolmTest(bool friedmanResult, bool tex)
+    int Statistics::getControlIdx()
+    {
+        if (!fitted) {
+            fit();
+        }
+        return controlIdx;
+    }
+    void Statistics::postHocTest()
+    {
+        if (score == "accuracy") {
+            postHocHolmTest();
+        } else {
+            postHocWilcoxonTest();
+        }
+    }
+    void Statistics::postHocWilcoxonTest()
+    {
+        if (!fitted) {
+            fit();
+        }
+        // Reference: Wilcoxon, F. (1945). “Individual Comparisons by Ranking Methods”. Biometrics Bulletin, 1(6), 80-83.
+        auto wilcoxon = WilcoxonTest(models, datasets, data, significance);
+        controlIdx = wilcoxon.getControlIdx();
+        postHocResults = wilcoxon.getPostHocResults();
+        setResultsOrder();
+        // Fill the ranks info
+        for (const auto& item : postHocResults) {
+            ranks[item.model] = item.rank;
+        }
+        Holm_Bonferroni();
+        restoreResultsOrder();
+    }
+    void Statistics::Holm_Bonferroni()
+    {
+        // The algorithm need the p-values sorted from the lowest to the highest
+        // Sort the models by p-value
+        std::sort(postHocResults.begin(), postHocResults.end(), [](const PostHocLine& a, const PostHocLine& b) {
+            return a.pvalue < b.pvalue;
+            });
+        // Holm adjustment
+        for (int i = 0; i < postHocResults.size(); ++i) {
+            auto item = postHocResults.at(i);
+            double before = i == 0 ? 0.0 : postHocResults.at(i - 1).pvalue;
+            double p_value = std::min((long double)1.0, item.pvalue * (nModels - i));
+            p_value = std::max(before, p_value);
+            postHocResults[i].pvalue = p_value;
+        }
+    }
+    void Statistics::setResultsOrder()
+    {
+        int c = 0;
+        for (auto& item : postHocResults) {
+            item.idx = c++;
+        }
+
+    }
+    void Statistics::restoreResultsOrder()
+    {
+        // Restore the order of the results
+        std::sort(postHocResults.begin(), postHocResults.end(), [](const PostHocLine& a, const PostHocLine& b) {
+            return a.idx < b.idx;
+            });
+    }
+    void Statistics::postHocHolmTest()
     {
         if (!fitted) {
             fit();
         }
-        std::stringstream oss;
         // Reference https://link.springer.com/article/10.1007/s44196-022-00083-8
         // Post-hoc Holm test
         // Calculate the p-value for the models paired with the control model
@@ -128,80 +211,66 @@ namespace platform {
         boost::math::normal dist(0.0, 1.0);
         double diff = sqrt(nModels * (nModels + 1) / (6.0 * nDatasets));
         for (int i = 0; i < nModels; i++) {
+            PostHocLine line;
+            line.model = models[i];
+            line.rank = ranks.at(models[i]);
+            line.wtl = wtl.at(i);
+            line.reject = false;
             if (i == controlIdx) {
-                stats[i] = 0.0;
+                postHocResults.push_back(line);
                 continue;
             }
             double z = std::abs(ranks.at(models[controlIdx]) - ranks.at(models[i])) / diff;
-            double p_value = (long double)2 * (1 - cdf(dist, z));
-            stats[i] = p_value;
+            line.pvalue = (long double)2 * (1 - cdf(dist, z));
+            line.reject = (line.pvalue < significance);
+            postHocResults.push_back(line);
         }
-        // Sort the models by p-value
-        std::vector<std::pair<int, double>> statsOrder;
-        for (const auto& stat : stats) {
-            statsOrder.push_back({ stat.first, stat.second });
-        }
-        std::sort(statsOrder.begin(), statsOrder.end(), [](const std::pair<int, double>& a, const std::pair<int, double>& b) {
-            return a.second < b.second;
+        std::sort(postHocResults.begin(), postHocResults.end(), [](const PostHocLine& a, const PostHocLine& b) {
+            return a.rank < b.rank;
             });
+        setResultsOrder();
+        Holm_Bonferroni();
+        restoreResultsOrder();
+    }
 
-        // Holm adjustment
-        for (int i = 0; i < statsOrder.size(); ++i) {
-            auto item = statsOrder.at(i);
-            double before = i == 0 ? 0.0 : statsOrder.at(i - 1).second;
-            double p_value = std::min((double)1.0, item.second * (nModels - i));
-            p_value = std::max(before, p_value);
-            statsOrder[i] = { item.first, p_value };
-        }
-        holmResult.model = models.at(controlIdx);
+    void Statistics::postHocTestReport(bool friedmanResult, bool tex)
+    {
+
+        std::stringstream oss;
         auto color = friedmanResult ? Colors::CYAN() : Colors::YELLOW();
         oss << color;
-        oss << "  *************************************************************************************************************" << std::endl;
-        oss << "  Post-hoc Holm test: H0: 'There is no significant differences between the control model and the other models.'" << std::endl;
+        oss << "  " << std::string(hlen + 25, '*') << std::endl;
+        oss << "  Post-hoc " << postHocType << " test: H0: 'There is no significant differences between the control model and the other models.'" << std::endl;
         oss << "  Control model: " << models.at(controlIdx) << std::endl;
         oss << "  " << std::left << std::setw(maxModelName) << std::string("Model") << " p-value      rank      win tie loss Status" << std::endl;
         oss << "  " << std::string(maxModelName, '=') << " ============ ========= === === ==== =============" << std::endl;
-        // sort ranks from lowest to highest
-        std::vector<std::pair<std::string, float>> ranksOrder;
-        for (const auto& rank : ranks) {
-            ranksOrder.push_back({ rank.first, rank.second });
-        }
-        std::sort(ranksOrder.begin(), ranksOrder.end(), [](const std::pair<std::string, float>& a, const std::pair<std::string, float>& b) {
-            return a.second < b.second;
-            });
-        // Show the control model info.
-        oss << "  " << Colors::BLUE() << std::left << std::setw(maxModelName) << ranksOrder.at(0).first << " ";
-        oss << std::setw(12) << " " << std::setprecision(7) << std::fixed << " " << ranksOrder.at(0).second << std::endl;
-        for (const auto& item : ranksOrder) {
-            auto idx = distance(models.begin(), find(models.begin(), models.end(), item.first));
-            double pvalue = 0.0;
-            for (const auto& stat : statsOrder) {
-                if (stat.first == idx) {
-                    pvalue = stat.second;
-                }
-            }
-            holmResult.holmLines.push_back({ item.first, pvalue, item.second, wtl.at(idx), pvalue < significance });
-            if (item.first == models.at(controlIdx)) {
+        bool first = true;
+        for (const auto& item : postHocResults) {
+            if (first) {
+                oss << "  " << Colors::BLUE() << std::left << std::setw(maxModelName) << item.model << " ";
+                oss << std::setw(12) << " " << std::setprecision(7) << std::fixed << " " << item.rank << std::endl;
+                first = false;
                 continue;
             }
+            auto pvalue = item.pvalue;
             auto colorStatus = pvalue > significance ? Colors::GREEN() : Colors::MAGENTA();
             auto status = pvalue > significance ? Symbols::check_mark : Symbols::cross;
             auto textStatus = pvalue > significance ? " accepted H0" : " rejected H0";
-            oss << "  " << colorStatus << std::left << std::setw(maxModelName) << item.first << " ";
-            oss << std::setprecision(6) << std::scientific << pvalue << std::setprecision(7) << std::fixed << " " << item.second;
-            oss << " " << std::right << std::setw(3) << wtl.at(idx).win << " " << std::setw(3) << wtl.at(idx).tie << " " << std::setw(4) << wtl.at(idx).loss;
+            oss << "  " << colorStatus << std::left << std::setw(maxModelName) << item.model << " ";
+            oss << std::setprecision(6) << std::scientific << pvalue << std::setprecision(7) << std::fixed << " " << item.rank;
+            oss << " " << std::right << std::setw(3) << item.wtl.win << " " << std::setw(3) << item.wtl.tie << " " << std::setw(4) << item.wtl.loss;
             oss << " " << status << textStatus << std::endl;
         }
-        oss << color << "  *************************************************************************************************************" << std::endl;
+        oss << color << "  " << std::string(hlen + 25, '*') << std::endl;
         oss << Colors::RESET();
         if (output) {
             std::cout << oss.str();
         }
         if (tex) {
-            BestResultsTex bestResultsTex;
+            BestResultsTex bestResultsTex(score);
             BestResultsMd bestResultsMd;
-            bestResultsTex.holm_test(holmResult, get_date() + " " + get_time());
-            bestResultsMd.holm_test(holmResult, get_date() + " " + get_time());
+            bestResultsTex.postHoc_test(postHocResults, postHocType, get_date() + " " + get_time());
+            bestResultsMd.postHoc_test(postHocResults, postHocType, get_date() + " " + get_time());
         }
     }
     bool Statistics::friedmanTest()
@@ -213,7 +282,7 @@ namespace platform {
         // Friedman test
         // Calculate the Friedman statistic
         oss << Colors::BLUE() << std::endl;
-        oss << "***************************************************************************************************************" << std::endl;
+        oss << std::string(hlen, '*') << std::endl;
         oss << Colors::GREEN() << "Friedman test: H0: 'There is no significant differences between all the classifiers.'" << Colors::BLUE() << std::endl;
         double degreesOfFreedom = nModels - 1.0;
         double sumSquared = 0;
@@ -238,23 +307,11 @@ namespace platform {
             oss << Colors::YELLOW() << "The null hypothesis H0 is accepted. Computed p-values will not be significant." << std::endl;
             result = false;
         }
-        oss << Colors::BLUE() << "***************************************************************************************************************" << Colors::RESET() << std::endl;
+        oss << Colors::BLUE() << std::string(hlen, '*') << Colors::RESET() << std::endl;
         if (output) {
             std::cout << oss.str();
         }
         friedmanResult = { friedmanQ, criticalValue, p_value, result };
         return result;
     }
-    FriedmanResult& Statistics::getFriedmanResult()
-    {
-        return friedmanResult;
-    }
-    HolmResult& Statistics::getHolmResult()
-    {
-        return holmResult;
-    }
-    std::map<std::string, std::map<std::string, float>>& Statistics::getRanks()
-    {
-        return ranksModels;
-    }
 } // namespace platform

src/best/Statistics.h

@@ -9,9 +9,9 @@ namespace platform {
     using json = nlohmann::ordered_json;
 
     struct WTL {
-        int win;
-        int tie;
-        int loss;
+        uint win;
+        uint tie;
+        uint loss;
     };
     struct FriedmanResult {
         double statistic;
@@ -19,29 +19,36 @@ namespace platform {
         long double pvalue;
         bool reject;
     };
-    struct HolmLine {
+    struct PostHocLine {
+        uint idx; //index of the main order
         std::string model;
         long double pvalue;
         double rank;
         WTL wtl;
         bool reject;
     };
-    struct HolmResult {
-        std::string model;
-        std::vector<HolmLine> holmLines;
-    };
+
     class Statistics {
     public:
-        Statistics(const std::vector<std::string>& models, const std::vector<std::string>& datasets, const json& data, double significance = 0.05, bool output = true);
+        Statistics(const std::string& score, const std::vector<std::string>& models, const std::vector<std::string>& datasets, const json& data, double significance = 0.05, bool output = true);
         bool friedmanTest();
-        void postHocHolmTest(bool friedmanResult, bool tex=false);
-        FriedmanResult& getFriedmanResult();
-        HolmResult& getHolmResult();
-        std::map<std::string, std::map<std::string, float>>& getRanks();
+        void postHocTest();
+        void postHocTestReport(bool friedmanResult, bool tex);
+        int getControlIdx();
+        FriedmanResult& getFriedmanResult() { return friedmanResult; }
+        std::vector<PostHocLine>& getPostHocResults() { return postHocResults; }
+        std::map<std::string, std::map<std::string, float>>& getRanks() { return ranksModels; } // ranks of the models per dataset
     private:
         void fit();
+        void postHocHolmTest();
+        void postHocWilcoxonTest();
         void computeRanks();
         void computeWTL();
+        void Holm_Bonferroni();
+        void setResultsOrder(); // Set the order of the results based on the statistic analysis needed
+        void restoreResultsOrder(); // Restore the order of the results after the Holm-Bonferroni adjustment
+        const std::string& score;
+        std::string postHocType;
         const std::vector<std::string>& models;
         const std::vector<std::string>& datasets;
         const json& data;
@@ -51,12 +58,14 @@ namespace platform {
         int nModels = 0;
         int nDatasets = 0;
         int controlIdx = 0;
+        int greaterAverage = -1; // The model with the greater average score
         std::map<int, WTL> wtl;
         std::map<std::string, float> ranks;
         int maxModelName = 0;
         int maxDatasetName = 0;
+        int hlen; // length of the line
         FriedmanResult friedmanResult;
-        HolmResult holmResult;
+        std::vector<PostHocLine> postHocResults;
         std::map<std::string, std::map<std::string, float>> ranksModels;
     };
 }

src/best/WilcoxonTest.hpp

@@ -0,0 +1,245 @@
+#ifndef BEST_WILCOXON_TEST_HPP
+#define BEST_WILCOXON_TEST_HPP
+// WilcoxonTest.hpp
+// Stand‑alone class for paired Wilcoxon signed‑rank post‑hoc analysis
+// ------------------------------------------------------------------
+//  * Constructor takes the *already‑loaded* nlohmann::json object plus the
+//    vectors of model and dataset names.
+//  * Internally selects a control model (highest average AUC) and builds all
+//    statistics (ranks, W/T/L counts, Wilcoxon p‑values).
+//  * Public API:
+//        int                  getControlIdx()      const;
+//        PostHocResult        getPostHocResult()   const;
+//
+#include <vector>
+#include <string>
+#include <cmath>
+#include <algorithm>
+#include <numeric>
+#include <limits>
+#include <nlohmann/json.hpp>
+#include "Statistics.h"
+
+namespace platform {
+    class WilcoxonTest {
+    public:
+        WilcoxonTest(const std::vector<std::string>& models, const std::vector<std::string>& datasets,
+            const json& data, double alpha = 0.05) : models_(models), datasets_(datasets), data_(data), alpha_(alpha)
+        {
+            buildAUCTable();              // extracts all AUCs into a dense matrix
+            computeAverageAUCs();         // per‑model mean (→ control selection)
+            computeAverageRanks();        // Friedman‑style ranks per model
+            selectControlModel();         // sets control_idx_
+            buildPostHocResult();         // fills postHocResult_
+        }
+
+        int getControlIdx() const noexcept { return control_idx_; }
+        const std::vector<PostHocLine>& getPostHocResults() const noexcept { return postHocResults_; }
+
+    private:
+        //-------------------------------------------------- helper structs ----
+        // When a value is missing we keep NaN so that ordinary arithmetic still
+        // works (NaN simply propagates and we can test with std::isnan).
+        using Matrix = std::vector<std::vector<double>>; // [model][dataset]
+
+        //------------------------------------------------- implementation ----
+        void buildAUCTable()
+        {
+            const std::size_t M = models_.size();
+            const std::size_t D = datasets_.size();
+            auc_.assign(M, std::vector<double>(D, std::numeric_limits<double>::quiet_NaN()));
+
+            for (std::size_t i = 0; i < M; ++i) {
+                const auto& model = models_[i];
+                for (std::size_t j = 0; j < D; ++j) {
+                    const auto& ds = datasets_[j];
+                    try {
+                        auc_[i][j] = data_.at(model).at(ds).at(0).get<double>();
+                    }
+                    catch (...) {
+                        // leave as NaN when value missing
+                    }
+                }
+            }
+        }
+
+        void computeAverageAUCs()
+        {
+            const std::size_t M = models_.size();
+            avg_auc_.resize(M, std::numeric_limits<double>::quiet_NaN());
+
+            for (std::size_t i = 0; i < M; ++i) {
+                double sum = 0.0;
+                std::size_t cnt = 0;
+                for (double v : auc_[i]) {
+                    if (!std::isnan(v)) { sum += v; ++cnt; }
+                }
+                avg_auc_[i] = cnt ? sum / cnt : std::numeric_limits<double>::quiet_NaN();
+            }
+        }
+
+        // Average rank across datasets (1 = best).
+        void computeAverageRanks()
+        {
+            const std::size_t M = models_.size();
+            const std::size_t D = datasets_.size();
+            rank_sum_.assign(M, 0.0);
+            rank_cnt_.assign(M, 0);
+
+            const double EPS = 1e-10;
+
+            for (std::size_t j = 0; j < D; ++j) {
+                // Collect present values for this dataset
+                std::vector<std::pair<double, std::size_t>> vals; // (auc, model_idx)
+                vals.reserve(M);
+                for (std::size_t i = 0; i < M; ++i) {
+                    if (!std::isnan(auc_[i][j]))
+                        vals.emplace_back(auc_[i][j], i);
+                }
+                if (vals.empty()) continue; // no info for this dataset
+
+                // Sort descending (higher AUC better)
+                std::sort(vals.begin(), vals.end(), [](auto a, auto b) {
+                    return a.first > b.first;
+                    });
+
+                // Assign ranks with average for ties
+                std::size_t k = 0;
+                while (k < vals.size()) {
+                    std::size_t l = k + 1;
+                    while (l < vals.size() && std::fabs(vals[l].first - vals[k].first) < EPS) ++l;
+                    const double avg_rank = (k + 1 + l) * 0.5; // average of ranks (1‑based)
+                    for (std::size_t m = k; m < l; ++m) {
+                        const auto idx = vals[m].second;
+                        rank_sum_[idx] += avg_rank;
+                        ++rank_cnt_[idx];
+                    }
+                    k = l;
+                }
+            }
+
+            // Final average
+            avg_rank_.resize(M, std::numeric_limits<double>::quiet_NaN());
+            for (std::size_t i = 0; i < M; ++i) {
+                avg_rank_[i] = rank_cnt_[i] ? rank_sum_[i] / rank_cnt_[i]
+                    : std::numeric_limits<double>::quiet_NaN();
+            }
+        }
+
+        void selectControlModel()
+        {
+            // pick model with highest average AUC (ties → first)
+            control_idx_ = 0;
+            for (std::size_t i = 1; i < avg_auc_.size(); ++i) {
+                if (avg_auc_[i] > avg_auc_[control_idx_]) control_idx_ = static_cast<int>(i);
+            }
+        }
+
+        void buildPostHocResult()
+        {
+            const std::size_t M = models_.size();
+            const std::size_t D = datasets_.size();
+            const std::string& control_name = models_[control_idx_];
+
+            const double practical_threshold = 0.0005; // same heuristic as original code
+
+            for (std::size_t i = 0; i < M; ++i) {
+                PostHocLine line;
+                line.model = models_[i];
+                line.rank = avg_auc_[i];
+
+                WTL wtl = { 0, 0, 0 }; // win, tie, loss
+                std::vector<double> differences;
+                differences.reserve(D);
+
+                for (std::size_t j = 0; j < D; ++j) {
+                    double auc_control = auc_[control_idx_][j];
+                    double auc_other = auc_[i][j];
+                    if (std::isnan(auc_control) || std::isnan(auc_other)) continue;
+
+                    double diff = auc_control - auc_other; // control − comparison
+                    if (std::fabs(diff) <= practical_threshold) {
+                        ++wtl.tie;
+                    } else if (diff < 0) {
+                        ++wtl.win;   // comparison wins
+                    } else {
+                        ++wtl.loss; // control wins
+                    }
+                    differences.push_back(diff);
+                }
+
+                line.wtl = wtl;
+                line.pvalue = differences.empty() ? 1.0L : static_cast<long double>(wilcoxonSignedRankTest(differences));
+                line.reject = (line.pvalue < alpha_);
+
+                postHocResults_.push_back(std::move(line));
+            }
+            // Sort results by rank (descending)
+            std::sort(postHocResults_.begin(), postHocResults_.end(), [](const PostHocLine& a, const PostHocLine& b) {
+                return a.rank > b.rank;
+                });
+        }
+
+        // ------------------------------------------------ Wilcoxon (private) --
+        static double wilcoxonSignedRankTest(const std::vector<double>& diffs)
+        {
+            if (diffs.empty()) return 1.0;
+
+            // Build |diff| + sign vector (exclude zeros)
+            struct Node { double absval; int sign; };
+            std::vector<Node> v;
+            v.reserve(diffs.size());
+            for (double d : diffs) {
+                if (d != 0.0) v.push_back({ std::fabs(d), d > 0 ? 1 : -1 });
+            }
+            if (v.empty()) return 1.0;
+
+            // Sort by absolute value
+            std::sort(v.begin(), v.end(), [](const Node& a, const Node& b) { return a.absval < b.absval; });
+
+            const double EPS = 1e-10;
+            const std::size_t n = v.size();
+            std::vector<double> ranks(n, 0.0);
+
+            std::size_t i = 0;
+            while (i < n) {
+                std::size_t j = i + 1;
+                while (j < n && std::fabs(v[j].absval - v[i].absval) < EPS) ++j;
+                double avg_rank = (i + 1 + j) * 0.5; // 1‑based ranks
+                for (std::size_t k = i; k < j; ++k) ranks[k] = avg_rank;
+                i = j;
+            }
+
+            double w_plus = 0.0, w_minus = 0.0;
+            for (std::size_t k = 0; k < n; ++k) {
+                if (v[k].sign > 0) w_plus += ranks[k];
+                else                w_minus += ranks[k];
+            }
+            double w = std::min(w_plus, w_minus);
+            double mean_w = n * (n + 1) / 4.0;
+            double sd_w = std::sqrt(n * (n + 1) * (2 * n + 1) / 24.0);
+            if (sd_w == 0.0) return 1.0; // degenerate (all diffs identical)
+
+            double z = (w - mean_w) / sd_w;
+            double p_two = std::erfc(std::fabs(z) / std::sqrt(2.0)); // 2‑sided tail
+            return p_two;
+        }
+
+        //-------------------------------------------------------- data ----
+        std::vector<std::string> models_;
+        std::vector<std::string> datasets_;
+        json                       data_;
+        double                     alpha_;
+
+        Matrix                     auc_;         // [model][dataset]
+        std::vector<double>        avg_auc_;     // mean AUC per model
+        std::vector<double>        avg_rank_;    // mean rank per model
+        std::vector<double>        rank_sum_;    // helper for ranks
+        std::vector<int>           rank_cnt_;    // datasets counted per model
+
+        int                        control_idx_ = -1;
+        std::vector<PostHocLine>   postHocResults_;
+    };
+
+} // namespace platform
+#endif // BEST_WILCOXON_TEST_HPP

src/commands/b_best.cpp

@@ -4,19 +4,22 @@
 #include "main/modelRegister.h"
 #include "common/Paths.h"
 #include "common/Colors.h"
+#include "common/Utils.h"
 #include "best/BestResults.h"
+#include "common/DotEnv.h"
 #include "config_platform.h"
 
 void manageArguments(argparse::ArgumentParser& program)
 {
-    program.add_argument("-m", "--model")
-        .help("Model to use or any")
-        .default_value("any");
+    auto env = platform::DotEnv();
+    program.add_argument("-m", "--model").help("Model to use or any").default_value("any");
+    program.add_argument("--folder").help("Results folder to use").default_value(platform::Paths::results());
     program.add_argument("-d", "--dataset").default_value("any").help("Filter results of the selected model) (any for all datasets)");
-    program.add_argument("-s", "--score").default_value("accuracy").help("Filter results of the score name supplied");
+    program.add_argument("-s", "--score").default_value(env.get("score")).help("Filter results of the score name supplied");
     program.add_argument("--friedman").help("Friedman test").default_value(false).implicit_value(true);
     program.add_argument("--excel").help("Output to excel").default_value(false).implicit_value(true);
-    program.add_argument("--tex").help("Output result table to TeX file").default_value(false).implicit_value(true);
+    program.add_argument("--tex").help("Output results to TeX & Markdown files").default_value(false).implicit_value(true);
+    program.add_argument("--index").help("In tex output show the index of the dataset instead of the name to save space").default_value(false).implicit_value(true);
     program.add_argument("--level").help("significance level").default_value(0.05).scan<'g', double>().action([](const std::string& value) {
         try {
             auto k = std::stod(value);
@@ -37,17 +40,22 @@ int main(int argc, char** argv)
 {
     argparse::ArgumentParser program("b_best", { platform_project_version.begin(), platform_project_version.end() });
     manageArguments(program);
-    std::string model, dataset, score;
-    bool build, report, friedman, excel, tex;
+    std::string model, dataset, score, folder;
+    bool build, report, friedman, excel, tex, index;
     double level;
     try {
         program.parse_args(argc, argv);
         model = program.get<std::string>("model");
+        folder = program.get<std::string>("folder");
+        if (folder.back() != '/') {
+            folder += '/';
+        }
         dataset = program.get<std::string>("dataset");
         score = program.get<std::string>("score");
         friedman = program.get<bool>("friedman");
         excel = program.get<bool>("excel");
         tex = program.get<bool>("tex");
+        index = program.get<bool>("index");
         level = program.get<double>("level");
         if (model == "" || score == "") {
             throw std::runtime_error("Model and score name must be supplied");
@@ -64,15 +72,20 @@ int main(int argc, char** argv)
         exit(1);
     }
     // Generate report
-    auto results = platform::BestResults(platform::Paths::results(), score, model, dataset, friedman, level);
+    auto results = platform::BestResults(folder, score, model, dataset, friedman, level);
     if (model == "any") {
         results.buildAll();
-        results.reportAll(excel, tex);
+        results.reportAll(excel, tex, index);
     } else {
         std::string fileName = results.build();
         std::cout << Colors::GREEN() << fileName << " created!" << Colors::RESET() << std::endl;
         results.reportSingle(excel);
     }
+    if (excel) {
+        auto fileName = results.getExcelFileName();
+        std::cout << "Opening " << fileName << std::endl;
+        platform::openFile(fileName);
+    }
     std::cout << Colors::RESET();
     return 0;
 }

src/commands/b_grid.cpp

@@ -1,16 +1,16 @@
 #include <iostream>
 #include <argparse/argparse.hpp>
 #include <map>
-#include <tuple>
 #include <nlohmann/json.hpp>
 #include <mpi.h>
 #include "main/Models.h"
-#include "main/modelRegister.h"
+#include "main/ArgumentsExperiment.h"
 #include "common/Paths.h"
-#include "common/Timer.h"
+#include "common/Timer.hpp"
 #include "common/Colors.h"
 #include "common/DotEnv.h"
 #include "grid/GridSearch.h"
+#include "grid/GridExperiment.h"
 #include "config_platform.h"
 
 using json = nlohmann::ordered_json;
@@ -31,15 +31,20 @@ void assignModel(argparse::ArgumentParser& parser)
             }
         );
 }
-void add_compute_args(argparse::ArgumentParser& program)
+void add_search_args(argparse::ArgumentParser& program)
 {
     auto env = platform::DotEnv();
     program.add_argument("--discretize").help("Discretize input datasets").default_value((bool)stoi(env.get("discretize"))).implicit_value(true);
     program.add_argument("--stratified").help("If Stratified KFold is to be done").default_value((bool)stoi(env.get("stratified"))).implicit_value(true);
     program.add_argument("--quiet").help("Don't display detailed progress").default_value(false).implicit_value(true);
     program.add_argument("--continue").help("Continue computing from that dataset").default_value(platform::GridSearch::NO_CONTINUE());
-    program.add_argument("--only").help("Used with continue to compute that dataset only").default_value(false).implicit_value(true);
+    program.add_argument("--only").help("Used with continue to search with that dataset only").default_value(false).implicit_value(true);
     program.add_argument("--exclude").default_value("[]").help("Datasets to exclude in json format, e.g. [\"dataset1\", \"dataset2\"]");
+    auto valid_choices = env.valid_tokens("smooth_strat");
+    auto& smooth_arg = program.add_argument("--smooth-strat").help("Smooth strategy used in Bayes Network node initialization. Valid values: " + env.valid_values("smooth_strat")).default_value(env.get("smooth_strat"));
+    for (auto choice : valid_choices) {
+        smooth_arg.choices(choice);
+    }
     program.add_argument("--nested").help("Set the double/nested cross validation number of folds").default_value(5).scan<'i', int>().action([](const std::string& value) {
         try {
             auto k = stoi(value);
@@ -133,7 +138,8 @@ void list_results(json& results, std::string& model)
     std::cout << std::string(MAXL, '*') << std::endl;
     int spaces = 7;
     int hyperparameters_spaces = 15;
-    for (const auto& item : results["results"].items()) {
+    nlohmann::json temp = results["results"]; // To show in alphabetical order of the dataset
+    for (const auto& item : temp.items()) {
         auto key = item.key();
         auto value = item.value();
         if (key.size() > spaces) {
@@ -148,7 +154,7 @@ void list_results(json& results, std::string& model)
     std::cout << "=== " << string(spaces, '=') << " " << string(19, '=') << " " << string(8, '=') << " "
         << string(8, '=') << " " << string(hyperparameters_spaces, '=') << std::endl;
     int index = 0;
-    for (const auto& item : results["results"].items()) {
+    for (const auto& item : temp.items()) {
         auto color = (index % 2) ? Colors::CYAN() : Colors::BLUE();
         auto value = item.value();
         std::cout << color;
@@ -181,13 +187,14 @@ void report(argparse::ArgumentParser& program)
         list_results(results, config.model);
     }
 }
-void compute(argparse::ArgumentParser& program)
+void search(argparse::ArgumentParser& program)
 {
     struct platform::ConfigGrid config;
     config.model = program.get<std::string>("model");
     config.score = program.get<std::string>("score");
     config.discretize = program.get<bool>("discretize");
     config.stratified = program.get<bool>("stratified");
+    config.smooth_strategy = program.get<std::string>("smooth-strat");
     config.n_folds = program.get<int>("folds");
     config.quiet = program.get<bool>("quiet");
     config.only = program.get<bool>("only");
@@ -199,9 +206,6 @@ void compute(argparse::ArgumentParser& program)
     }
     auto excluded = program.get<std::string>("exclude");
     config.excluded = json::parse(excluded);
-
-    auto env = platform::DotEnv();
-    config.platform = env.get("platform");
     platform::Paths::createPath(platform::Paths::grid());
     auto grid_search = platform::GridSearch(config);
     platform::Timer timer;
@@ -212,16 +216,48 @@ void compute(argparse::ArgumentParser& program)
     MPI_Comm_rank(MPI_COMM_WORLD, &mpi_config.rank);
     MPI_Comm_size(MPI_COMM_WORLD, &mpi_config.n_procs);
     if (mpi_config.n_procs < 2) {
-        throw std::runtime_error("Cannot use --compute with less than 2 mpi processes, try mpirun -np 2 ...");
+        throw std::runtime_error("Cannot use --search with less than 2 mpi processes, try mpirun -np 2 ...");
     }
     grid_search.go(mpi_config);
     if (mpi_config.rank == mpi_config.manager) {
         auto results = grid_search.loadResults();
+        std::cout << Colors::RESET() << "* Report of the computed hyperparameters" << std::endl;
         list_results(results, config.model);
         std::cout << "Process took " << timer.getDurationString() << std::endl;
     }
     MPI_Finalize();
 }
+void experiment(argparse::ArgumentParser& program)
+{
+    struct platform::ConfigGrid config;
+    auto arguments = platform::ArgumentsExperiment(program, platform::experiment_t::GRID);
+    arguments.parse();
+    auto path_results = arguments.getPathResults();
+    auto grid_experiment = platform::GridExperiment(arguments, config);
+    platform::Timer timer;
+    timer.start();
+    struct platform::ConfigMPI mpi_config;
+    mpi_config.manager = 0; // which process is the manager
+    MPI_Init(nullptr, nullptr);
+    MPI_Comm_rank(MPI_COMM_WORLD, &mpi_config.rank);
+    MPI_Comm_size(MPI_COMM_WORLD, &mpi_config.n_procs);
+    if (mpi_config.n_procs < 2) {
+        throw std::runtime_error("Cannot use --experiment with less than 2 mpi processes, try mpirun -np 2 ...");
+    }
+    grid_experiment.go(mpi_config);
+    if (mpi_config.rank == mpi_config.manager) {
+        auto experiment = grid_experiment.getExperiment();
+        std::cout << "* Report of the computed hyperparameters" << std::endl;
+        auto duration = timer.getDuration();
+        experiment.setDuration(duration);
+        if (grid_experiment.haveToSaveResults()) {
+            experiment.saveResult(path_results);
+        }
+        experiment.report();
+        std::cout << "Process took " << duration << std::endl;
+    }
+    MPI_Finalize();
+}
 int main(int argc, char** argv)
 {
     //
@@ -238,15 +274,21 @@ int main(int argc, char** argv)
     assignModel(report_command);
     report_command.add_description("Report the computed hyperparameters of a model.");
 
-    // grid compute subparser
-    argparse::ArgumentParser compute_command("compute");
-    compute_command.add_description("Compute using mpi the hyperparameters of a model.");
-    assignModel(compute_command);
-    add_compute_args(compute_command);
+    // grid search subparser
+    argparse::ArgumentParser search_command("search");
+    search_command.add_description("Search using mpi the hyperparameters of a model.");
+    assignModel(search_command);
+    add_search_args(search_command);
 
+    // grid experiment subparser
+    argparse::ArgumentParser experiment_command("experiment");
+    experiment_command.add_description("Experiment like b_main using mpi.");
+    auto arguments = platform::ArgumentsExperiment(experiment_command, platform::experiment_t::GRID);
+    arguments.add_arguments();
     program.add_subparser(dump_command);
     program.add_subparser(report_command);
-    program.add_subparser(compute_command);
+    program.add_subparser(search_command);
+    program.add_subparser(experiment_command);
 
     // 
     // Process options
@@ -254,7 +296,7 @@ int main(int argc, char** argv)
     try {
         program.parse_args(argc, argv);
         bool found = false;
-        map<std::string, void(*)(argparse::ArgumentParser&)> commands = { {"dump", &dump}, {"report", &report}, {"compute", &compute} };
+        map<std::string, void(*)(argparse::ArgumentParser&)> commands = { {"dump", &dump}, {"report", &report}, {"search", &search}, { "experiment",&experiment } };
         for (const auto& command : commands) {
             if (program.is_subcommand_used(command.first)) {
                 std::invoke(command.second, program.at<argparse::ArgumentParser>(command.first));
@@ -263,7 +305,7 @@ int main(int argc, char** argv)
             }
         }
         if (!found) {
-            throw std::runtime_error("You must specify one of the following commands: dump, report, compute\n");
+            throw std::runtime_error("You must specify one of the following commands: dump, experiment, report, search \n");
         }
     }
     catch (const exception& err) {

src/commands/b_list.cpp

@@ -8,6 +8,7 @@
 #include "common/Paths.h"
 #include "common/Colors.h"
 #include "common/Datasets.h"
+#include "common/Utils.h"
 #include "reports/DatasetsExcel.h"
 #include "reports/DatasetsConsole.h"
 #include "results/ResultsDatasetConsole.h"
@@ -24,9 +25,13 @@ void list_datasets(argparse::ArgumentParser& program)
     std::cout << report.getOutput();
     if (excel) {
         auto data = report.getData();
-        auto report = platform::DatasetsExcel();
-        report.report(data);
-        std::cout << std::endl << Colors::GREEN() << "Output saved in " << report.getFileName() << std::endl;
+        auto ereport = new platform::DatasetsExcel();
+        ereport->report(data);
+        std::cout << std::endl << Colors::GREEN() << "Output saved in " << ereport->getFileName() << std::endl;
+        auto fileName = ereport->getExcelFileName();
+        delete ereport;
+        std::cout << "Opening " << fileName << std::endl;
+        platform::openFile(fileName);
     }
 }
 
@@ -42,9 +47,13 @@ void list_results(argparse::ArgumentParser& program)
     std::cout << report.getOutput();
     if (excel) {
         auto data = report.getData();
-        auto report = platform::ResultsDatasetExcel();
-        report.report(data);
-        std::cout << std::endl << Colors::GREEN() << "Output saved in " << report.getFileName() << std::endl;
+        auto ereport = new platform::ResultsDatasetExcel();
+        ereport->report(data);
+        std::cout << std::endl << Colors::GREEN() << "Output saved in " << ereport->getFileName() << std::endl;
+        auto fileName = ereport->getExcelFileName();
+        delete ereport;
+        std::cout << "Opening " << fileName << std::endl;
+        platform::openFile(fileName);
     }
 }
 

src/commands/b_main.cpp

@@ -1,234 +1,37 @@
-#include <iostream>
 #include <argparse/argparse.hpp>
-#include <nlohmann/json.hpp>
 #include "main/Experiment.h"
-#include "common/Datasets.h"
-#include "common/DotEnv.h"
-#include "common/Paths.h"
-#include "main/Models.h"
-#include "main/modelRegister.h"
+#include "main/ArgumentsExperiment.h"
 #include "config_platform.h"
 
 
 using json = nlohmann::ordered_json;
 
 
-void manageArguments(argparse::ArgumentParser& program)
-{
-    auto env = platform::DotEnv();
-    auto datasets = platform::Datasets(false, platform::Paths::datasets());
-    auto& group = program.add_mutually_exclusive_group(true);
-    group.add_argument("-d", "--dataset")
-        .help("Dataset file name: " + datasets.toString())
-        .default_value("all")
-        .action([](const std::string& value) {
-        auto datasets = platform::Datasets(false, platform::Paths::datasets());
-        static std::vector<std::string> choices_datasets(datasets.getNames());
-        choices_datasets.push_back("all");
-        if (find(choices_datasets.begin(), choices_datasets.end(), value) != choices_datasets.end()) {
-            return value;
-        }
-        throw std::runtime_error("Dataset must be one of: " + datasets.toString());
-            }
-        );
-    group.add_argument("--datasets").nargs(1, 50).help("Datasets file names 1..50 separated by spaces").default_value(std::vector<std::string>());
-    group.add_argument("--datasets-file").default_value("").help("Datasets file name. Mutually exclusive with dataset. This file should contain a list of datasets to test.");
-    program.add_argument("--hyperparameters").default_value("{}").help("Hyperparameters passed to the model in Experiment");
-    program.add_argument("--hyper-file").default_value("").help("Hyperparameters file name." \
-        "Mutually exclusive with hyperparameters. This file should contain hyperparameters for each dataset in json format.");
-    program.add_argument("--hyper-best").default_value(false).help("Use best results of the model as source of hyperparameters").implicit_value(true);
-    program.add_argument("-m", "--model")
-        .help("Model to use: " + platform::Models::instance()->toString())
-        .action([](const std::string& value) {
-        static const std::vector<std::string> choices = platform::Models::instance()->getNames();
-        if (find(choices.begin(), choices.end(), value) != choices.end()) {
-            return value;
-        }
-        throw std::runtime_error("Model must be one of " + platform::Models::instance()->toString());
-            }
-        );
-    program.add_argument("--title").default_value("").help("Experiment title");
-    program.add_argument("--discretize").help("Discretize input dataset").default_value((bool)stoi(env.get("discretize"))).implicit_value(true);
-    auto valid_choices = env.valid_tokens("discretize_algo");
-    auto& disc_arg = program.add_argument("--discretize-algo").help("Algorithm to use in discretization. Valid values: " + env.valid_values("discretize_algo")).default_value(env.get("discretize_algo"));
-    for (auto choice : valid_choices) {
-        disc_arg.choices(choice);
-    }
-    valid_choices = env.valid_tokens("smooth_strat");
-    auto& smooth_arg = program.add_argument("--smooth-strat").help("Smooth strategy used in Bayes Network node initialization. Valid values: " + env.valid_values("smooth_strat")).default_value(env.get("smooth_strat"));
-    for (auto choice : valid_choices) {
-        smooth_arg.choices(choice);
-    }
-    auto& score_arg = program.add_argument("-s", "--score").help("Score to use. Valid values: " + env.valid_values("score")).default_value(env.get("score"));
-    valid_choices = env.valid_tokens("score");
-    for (auto choice : valid_choices) {
-        score_arg.choices(choice);
-    }
-    program.add_argument("--generate-fold-files").help("generate fold information in datasets_experiment folder").default_value(false).implicit_value(true);
-    program.add_argument("--graph").help("generate graphviz dot files with the model").default_value(false).implicit_value(true);
-    program.add_argument("--no-train-score").help("Don't compute train score").default_value(false).implicit_value(true);
-    program.add_argument("--quiet").help("Don't display detailed progress").default_value(false).implicit_value(true);
-    program.add_argument("--save").help("Save result (always save if no dataset is supplied)").default_value(false).implicit_value(true);
-    program.add_argument("--stratified").help("If Stratified KFold is to be done").default_value((bool)stoi(env.get("stratified"))).implicit_value(true);
-    program.add_argument("-f", "--folds").help("Number of folds").default_value(stoi(env.get("n_folds"))).scan<'i', int>().action([](const std::string& value) {
-        try {
-            auto k = stoi(value);
-            if (k < 2) {
-                throw std::runtime_error("Number of folds must be greater than 1");
-            }
-            return k;
-        }
-        catch (const runtime_error& err) {
-            throw std::runtime_error(err.what());
-        }
-        catch (...) {
-            throw std::runtime_error("Number of folds must be an integer");
-        }});
-        auto seed_values = env.getSeeds();
-        program.add_argument("--seeds").nargs(1, 10).help("Random seeds. Set to -1 to have pseudo random").scan<'i', int>().default_value(seed_values);
-}
-
 int main(int argc, char** argv)
 {
     argparse::ArgumentParser program("b_main", { platform_project_version.begin(), platform_project_version.end() });
-    manageArguments(program);
-    std::string file_name, model_name, title, hyperparameters_file, datasets_file, discretize_algo, smooth_strat, score;
-    json hyperparameters_json;
-    bool discretize_dataset, stratified, saveResults, quiet, no_train_score, generate_fold_files, graph, hyper_best;
-    std::vector<int> seeds;
-    std::vector<std::string> file_names;
-    std::vector<std::string> filesToTest;
-    int n_folds;
-    try {
-        program.parse_args(argc, argv);
-        file_name = program.get<std::string>("dataset");
-        file_names = program.get<std::vector<std::string>>("datasets");
-        datasets_file = program.get<std::string>("datasets-file");
-        model_name = program.get<std::string>("model");
-        discretize_dataset = program.get<bool>("discretize");
-        discretize_algo = program.get<std::string>("discretize-algo");
-        smooth_strat = program.get<std::string>("smooth-strat");
-        stratified = program.get<bool>("stratified");
-        quiet = program.get<bool>("quiet");
-        graph = program.get<bool>("graph");
-        n_folds = program.get<int>("folds");
-        score = program.get<std::string>("score");
-        seeds = program.get<std::vector<int>>("seeds");
-        auto hyperparameters = program.get<std::string>("hyperparameters");
-        hyperparameters_json = json::parse(hyperparameters);
-        hyperparameters_file = program.get<std::string>("hyper-file");
-        no_train_score = program.get<bool>("no-train-score");
-        hyper_best = program.get<bool>("hyper-best");
-        generate_fold_files = program.get<bool>("generate-fold-files");
-        if (hyper_best) {
-            // Build the best results file_name
-            hyperparameters_file = platform::Paths::results() + platform::Paths::bestResultsFile(score, model_name);
-            // ignore this parameter
-            hyperparameters = "{}";
-        } else {
-            if (hyperparameters_file != "" && hyperparameters != "{}") {
-                throw runtime_error("hyperparameters and hyper_file are mutually exclusive");
-            }
-        }
-        title = program.get<std::string>("title");
-        if (title == "" && file_name == "all") {
-            throw runtime_error("title is mandatory if all datasets are to be tested");
-        }
-        saveResults = program.get<bool>("save");
-    }
-    catch (const exception& err) {
-        cerr << err.what() << std::endl;
-        cerr << program;
-        exit(1);
-    }
-    auto datasets = platform::Datasets(false, platform::Paths::datasets());
-    if (datasets_file != "") {
-        ifstream catalog(datasets_file);
-        if (catalog.is_open()) {
-            std::string line;
-            while (getline(catalog, line)) {
-                if (line.empty() || line[0] == '#') {
-                    continue;
-                }
-                if (!datasets.isDataset(line)) {
-                    cerr << "Dataset " << line << " not found" << std::endl;
-                    exit(1);
-                }
-                filesToTest.push_back(line);
-            }
-            catalog.close();
-            saveResults = true;
-            if (title == "") {
-                title = "Test " + to_string(filesToTest.size()) + " datasets (" + datasets_file + ") "\
-                    + model_name + " " + to_string(n_folds) + " folds";
-            }
-        } else {
-            throw std::invalid_argument("Unable to open catalog file. [" + datasets_file + "]");
-        }
-    } else {
-        if (file_names.size() > 0) {
-            for (auto file : file_names) {
-                if (!datasets.isDataset(file)) {
-                    cerr << "Dataset " << file << " not found" << std::endl;
-                    exit(1);
-                }
-            }
-            filesToTest = file_names;
-            saveResults = true;
-            if (title == "") {
-                title = "Test " + to_string(file_names.size()) + " datasets " + model_name + " " + to_string(n_folds) + " folds";
-            }
-        } else {
-            if (file_name != "all") {
-                if (!datasets.isDataset(file_name)) {
-                    cerr << "Dataset " << file_name << " not found" << std::endl;
-                    exit(1);
-                }
-                if (title == "") {
-                    title = "Test " + file_name + " " + model_name + " " + to_string(n_folds) + " folds";
-                }
-                filesToTest.push_back(file_name);
-            } else {
-                filesToTest = datasets.getNames();
-                saveResults = true;
-            }
-        }
-    }
-
-    platform::HyperParameters test_hyperparams;
-    if (hyperparameters_file != "") {
-        test_hyperparams = platform::HyperParameters(datasets.getNames(), hyperparameters_file, hyper_best);
-    } else {
-        test_hyperparams = platform::HyperParameters(datasets.getNames(), hyperparameters_json);
-    }
-
+    auto arguments = platform::ArgumentsExperiment(program, platform::experiment_t::NORMAL);
+    arguments.add_arguments();
+    arguments.parse_args(argc, argv);
     /*
      * Begin Processing
      */
-    auto env = platform::DotEnv();
-    auto experiment = platform::Experiment();
-    experiment.setTitle(title).setLanguage("c++").setLanguageVersion("gcc 14.1.1");
-    experiment.setDiscretizationAlgorithm(discretize_algo).setSmoothSrategy(smooth_strat);
-    experiment.setDiscretized(discretize_dataset).setModel(model_name).setPlatform(env.get("platform"));
-    experiment.setStratified(stratified).setNFolds(n_folds).setScoreName(score);
-    experiment.setHyperparameters(test_hyperparams);
-    for (auto seed : seeds) {
-        experiment.addRandomSeed(seed);
-    }
+     // Initialize the experiment class with the command line arguments
+    auto experiment = arguments.initializedExperiment();
+    auto path_results = arguments.getPathResults();
     platform::Timer timer;
     timer.start();
-    experiment.go(filesToTest, quiet, no_train_score, generate_fold_files, graph);
+    experiment.go();
     experiment.setDuration(timer.getDuration());
-    if (!quiet) {
+    if (!arguments.isQuiet()) {
         // Classification report if only one dataset is tested
-        experiment.report(filesToTest.size() == 1);
+        experiment.report();
     }
-    if (saveResults) {
-        experiment.saveResult();
+    if (arguments.haveToSaveResults()) {
+        experiment.saveResult(path_results);
     }
-    if (graph) {
+    if (arguments.doGraph()) {
         experiment.saveGraph();
     }
-    std::cout << "Done!" << std::endl;
     return 0;
 }

src/commands/b_manage.cpp

@@ -1,7 +1,8 @@
+
+#include <utility>
 #include <iostream>
 #include <sys/ioctl.h>
-#include <utility>
-#include <unistd.h>
+#include "common/Paths.h"
 #include <argparse/argparse.hpp>
 #include "manage/ManageScreen.h"
 #include <signal.h>
@@ -13,6 +14,7 @@ void manageArguments(argparse::ArgumentParser& program, int argc, char** argv)
 {
     program.add_argument("-m", "--model").default_value("any").help("Filter results of the selected model)");
     program.add_argument("-s", "--score").default_value("any").help("Filter results of the score name supplied");
+    program.add_argument("--folder").help("Results folder to use").default_value(platform::Paths::results());
     program.add_argument("--platform").default_value("any").help("Filter results of the selected platform");
     program.add_argument("--complete").help("Show only results with all datasets").default_value(false).implicit_value(true);
     program.add_argument("--partial").help("Show only partial results").default_value(false).implicit_value(true);
@@ -51,11 +53,17 @@ void handleResize(int sig)
     manager->updateSize(rows, cols);
 }
 
+
+
 int main(int argc, char** argv)
 {
     auto program = argparse::ArgumentParser("b_manage", { platform_project_version.begin(), platform_project_version.end() });
     manageArguments(program, argc, argv);
     std::string model = program.get<std::string>("model");
+    std::string path = program.get<std::string>("folder");
+    if (path.back() != '/') {
+        path += '/';
+    }
     std::string score = program.get<std::string>("score");
     std::string platform = program.get<std::string>("platform");
     bool complete = program.get<bool>("complete");
@@ -65,8 +73,13 @@ int main(int argc, char** argv)
         partial = false;
     signal(SIGWINCH, handleResize);
     auto [rows, cols] = numRowsCols();
-    manager = new platform::ManageScreen(rows, cols, model, score, platform, complete, partial, compare);
+    manager = new platform::ManageScreen(path, rows, cols, model, score, platform, complete, partial, compare);
     manager->doMenu();
+    auto fileName = manager->getExcelFileName();
     delete manager;
+    if (!fileName.empty()) {
+        std::cout << "Opening " << fileName << std::endl;
+        platform::openFile(fileName);
+    }
     return 0;
 }

src/commands/b_results.cpp

@@ -0,0 +1,102 @@
+#include <iostream>
+#include <filesystem>
+#include <fstream>
+#include <vector>
+#include "nlohmann/json.hpp"
+#include "argparse/argparse.hpp"
+#include "common/Paths.h"
+#include "results/JsonValidator.h"
+#include "results/SchemaV1_0.h"
+#include "config_platform.h"
+
+using json = nlohmann::json;
+namespace fs = std::filesystem;
+void header(const std::string& message, int length, const std::string& symbol)
+{
+    std::cout << std::string(length + 11, symbol[0]) << std::endl;
+    std::cout << symbol << " " << std::setw(length + 7) << std::left << message << " " << symbol << std::endl;
+    std::cout << std::string(length + 11, symbol[0]) << std::endl;
+}
+int main(int argc, char* argv[])
+{
+    argparse::ArgumentParser program("b_results", { platform_project_version.begin(), platform_project_version.end() });
+    program.add_description("Check the results files and optionally fixes them.");
+    program.add_argument("--fix").help("Fix any errors in results").default_value(false).implicit_value(true);
+    program.add_argument("--file").help("check only this results file").default_value("");
+    std::string nameSuffix = "results_";
+    std::string schemaVersion = "1.0";
+    bool fix_it = false;
+    std::string selected_file;
+    try {
+        program.parse_args(argc, argv);
+        fix_it = program.get<bool>("fix");
+        selected_file = program.get<std::string>("file");
+    }
+    catch (const std::exception& err) {
+        std::cerr << err.what() << std::endl;
+        std::cerr << program;
+        exit(1);
+    }
+    //
+    // Determine the files to process
+    //
+    std::vector<std::string> result_files;
+    int max_length = 0;
+    if (selected_file != "") {
+        if (!selected_file.starts_with(platform::Paths::results())) {
+            selected_file = platform::Paths::results() + selected_file;
+        }
+        // Only check the selected file
+        result_files.push_back(selected_file);
+        max_length = selected_file.length();
+    } else {
+        // Load the result files and find the longest file name
+        for (const auto& entry : fs::directory_iterator(platform::Paths::results())) {
+            if (entry.is_regular_file() && entry.path().filename().string().starts_with(nameSuffix) && entry.path().filename().string().ends_with(".json")) {
+                std::string fileName = entry.path().string();
+                if (fileName.length() > max_length) {
+                    max_length = fileName.length();
+                }
+                result_files.push_back(fileName);
+            }
+        }
+    }
+    //
+    // Process the results files
+    //
+    if (result_files.empty()) {
+        std::cerr << "Error: No result files found." << std::endl;
+        return 1;
+    }
+    std::string header_message = "Processing " + std::to_string(result_files.size()) + " result files.";
+    header(header_message, max_length, "*");
+    platform::JsonValidator validator(platform::SchemaV1_0::schema);
+    int n_errors = 0;
+    std::vector<std::string> files_with_errors;
+    for (const auto& file_name : result_files) {
+        std::vector<std::string> errors = validator.validate_file(file_name);
+        if (!errors.empty()) {
+            n_errors++;
+            std::cout << std::setw(max_length) << std::left << file_name << ": " << errors.size() << " Errors:" << std::endl;
+            for (const auto& error : errors) {
+                std::cout << " - " << error << std::endl;
+            }
+            if (fix_it) {
+                validator.fix_it(file_name);
+                std::cout << " -> File fixed." << std::endl;
+            }
+            files_with_errors.push_back(file_name);
+        }
+    }
+    if (n_errors == 0) {
+        header("All files are valid.", max_length, "*");
+    } else {
+        std::string $verb = (fix_it) ? "had" : "have";
+        std::string msg = std::to_string(n_errors) + " files " + $verb + " errors.";
+        header(msg, max_length, "*");
+        for (const auto& file_name : files_with_errors) {
+            std::cout << "- " << file_name << std::endl;
+        }
+    }
+    return 0;
+}

src/common/Datasets.cpp

@@ -1,4 +1,5 @@
 #include <fstream>
+#include<algorithm>
 #include "Datasets.h"
 #include <nlohmann/json.hpp>
 
@@ -24,10 +25,20 @@ namespace platform {
             throw std::invalid_argument("Unable to open catalog file. [" + path + "all.txt" + "]");
         }
         std::string line;
+        std::vector<std::string> sorted_lines;
         while (getline(catalog, line)) {
             if (line.empty() || line[0] == '#') {
                 continue;
             }
+            sorted_lines.push_back(line);
+        }
+        sort(sorted_lines.begin(), sorted_lines.end(), [](const auto& lhs, const auto& rhs) {
+            const auto result = mismatch(lhs.cbegin(), lhs.cend(), rhs.cbegin(), rhs.cend(), [](const auto& lhs, const auto& rhs) {return tolower(lhs) == tolower(rhs);});
+
+            return result.second != rhs.cend() && (result.first == lhs.cend() || tolower(*result.first) < tolower(*result.second));
+            });
+
+        for (const auto& line : sorted_lines) {
             std::vector<std::string> tokens = split(line, ';');
             std::string name = tokens[0];
             std::string className;
@@ -70,6 +81,11 @@ namespace platform {
     {
         std::vector<std::string> result;
         transform(datasets.begin(), datasets.end(), back_inserter(result), [](const auto& d) { return d.first; });
+        sort(result.begin(), result.end(), [](const auto& lhs, const auto& rhs) {
+            const auto result = mismatch(lhs.cbegin(), lhs.cend(), rhs.cbegin(), rhs.cend(), [](const auto& lhs, const auto& rhs) {return tolower(lhs) == tolower(rhs);});
+
+            return result.second != rhs.cend() && (result.first == lhs.cend() || tolower(*result.first) < tolower(*result.second));
+            });
         return result;
     }
     bool Datasets::isDataset(const std::string& name) const

src/common/Paths.h

@@ -49,6 +49,7 @@ namespace platform {
             return "BestResults_" + score + ".xlsx";
         }
         static std::string excelResults() { return "some_results.xlsx"; }
+        static std::string excelDatasets() { return "datasets.xlsx"; }
         static std::string grid_input(const std::string& model)
         {
             return grid() + "grid_" + model + "_input.json";
@@ -73,6 +74,7 @@ namespace platform {
         {
             return "post_hoc.md";
         }
+
     };
 }
 #endif

src/common/TensorUtils.h

@@ -0,0 +1,18 @@
+#ifndef TENSOR_UTILS_H
+#define TENSOR_UTILS_H
+
+#include <torch/torch.h>
+#include <vector>
+
+namespace platform {
+    template <typename T>
+    std::vector<T> tensorToVector(const torch::Tensor& tensor)
+    {
+        torch::Tensor contig_tensor = tensor.contiguous();
+        auto num_elements = contig_tensor.numel();
+        const T* tensor_data = contig_tensor.data_ptr<T>();
+        std::vector<T> result(tensor_data, tensor_data + num_elements);
+        return result;
+    }
+}
+#endif

src/common/Utils.h

@@ -1,20 +1,20 @@
 #ifndef UTILS_H
 #define UTILS_H
+
+#include <unistd.h>
 #include <sstream>
 #include <string>
 #include <vector>
 #include <algorithm>
-#include <torch/torch.h>
+#include <cstdlib>
+#include <cmath>
+#include <ctime>
+#include <iomanip>
+#include <string.h>
+
+extern char** environ;
+
 namespace platform {
-    template <typename T>
-    std::vector<T> tensorToVector(const torch::Tensor& tensor)
-    {
-        torch::Tensor contig_tensor = tensor.contiguous();
-        auto num_elements = contig_tensor.numel();
-        const T* tensor_data = contig_tensor.data_ptr<T>();
-        std::vector<T> result(tensor_data, tensor_data + num_elements);
-        return result;
-    }
     static std::string trim(const std::string& str)
     {
         std::string result = str;
@@ -66,5 +66,64 @@ namespace platform {
         oss << std::put_time(timeinfo, "%H:%M:%S");
         return oss.str();
     }
+    static void openFile(const std::string& fileName)
+    {
+        // #ifdef __APPLE__
+        //     // macOS uses the "open" command
+        //     std::string command = "open";
+        // #elif defined(__linux__)
+        //     // Linux typically uses "xdg-open"
+        //     std::string command = "xdg-open";
+        // #else
+        //     // For other OSes, do nothing or handle differently
+        //     std::cerr << "Unsupported platform." << std::endl;
+        //     return;
+        // #endif
+        //     execlp(command.c_str(), command.c_str(), fileName.c_str(), NULL);
+#ifdef __APPLE__
+        const char* tool = "/usr/bin/open";
+#elif defined(__linux__)
+        const char* tool = "/usr/bin/xdg-open";
+#else
+        std::cerr << "Unsupported platform." << std::endl;
+        return;
+#endif
+
+        // We'll build an argv array for execve:
+        std::vector<char*> argv;
+        argv.push_back(const_cast<char*>(tool));               // argv[0]
+        argv.push_back(const_cast<char*>(fileName.c_str()));   // argv[1]
+        argv.push_back(nullptr);
+
+        // Make a new environment array, skipping BASH_FUNC_ variables
+        std::vector<std::string> filteredEnv;
+        for (char** env = environ; *env != nullptr; ++env) {
+            // *env is a string like "NAME=VALUE"
+            // We want to skip those starting with "BASH_FUNC_"
+            if (strncmp(*env, "BASH_FUNC_", 10) == 0) {
+                // skip it
+                continue;
+            }
+            filteredEnv.push_back(*env);
+        }
+
+        // Convert filteredEnv into a char* array
+        std::vector<char*> envp;
+        for (auto& var : filteredEnv) {
+            envp.push_back(const_cast<char*>(var.c_str()));
+        }
+        envp.push_back(nullptr);
+
+        // Now call execve with the cleaned environment
+        // NOTE: You may need a full path to the tool if it's not in PATH, or use which() logic
+        // For now, let's assume "open" or "xdg-open" is found in the default PATH:
+        execve(tool, argv.data(), envp.data());
+
+        // If we reach here, execve failed
+        perror("execve failed");
+        // This would terminate your current process if it's not in a child
+        // Usually you'd do something like:
+        _exit(EXIT_FAILURE);
+    }
 }
 #endif

src/experimental_clfs/AdaBoost.cpp

@@ -0,0 +1,492 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include "AdaBoost.h"
+#include "DecisionTree.h"
+#include <cmath>
+#include <algorithm>
+#include <numeric>
+#include <sstream>
+#include <iomanip>
+#include "TensorUtils.hpp"
+
+// Conditional debug macro for performance-critical sections
+#define DEBUG_LOG(condition, ...) \
+    do { \
+        if (__builtin_expect((condition), 0)) { \
+            std::cout << __VA_ARGS__ << std::endl; \
+        } \
+    } while(0)
+
+namespace bayesnet {
+
+    AdaBoost::AdaBoost(int n_estimators, int max_depth)
+        : Ensemble(true), n_estimators(n_estimators), base_max_depth(max_depth), n(0), n_classes(0)
+    {
+        validHyperparameters = { "n_estimators", "base_max_depth" };
+    }
+
+    // Versión optimizada de buildModel - Reemplazar en AdaBoost.cpp:
+
+    void AdaBoost::buildModel(const torch::Tensor& weights)
+    {
+        // Initialize variables
+        models.clear();
+        alphas.clear();
+        training_errors.clear();
+
+        // Initialize n (number of features) and n_classes
+        n = dataset.size(0) - 1;  // Exclude the label row
+        n_classes = states[className].size();
+
+        // Initialize sample weights uniformly
+        int n_samples = dataset.size(1);
+        sample_weights = torch::ones({ n_samples }) / n_samples;
+
+        // If initial weights are provided, incorporate them
+        if (weights.defined() && weights.numel() > 0) {
+            if (weights.size(0) != n_samples) {
+                throw std::runtime_error("weights must have the same length as number of samples");
+            }
+            sample_weights = weights.clone();
+            normalizeWeights();
+        }
+
+        // Conditional debug information (only when debug is enabled)
+        DEBUG_LOG(debug, "Starting AdaBoost training with " << n_estimators << " estimators\n"
+            << "Number of classes: " << n_classes << "\n"
+            << "Number of features: " << n << "\n"
+            << "Number of samples: " << n_samples);
+
+        // Pre-compute random guess error threshold
+        const double random_guess_error = 1.0 - (1.0 / static_cast<double>(n_classes));
+
+        // Main AdaBoost training loop (SAMME algorithm)
+        for (int iter = 0; iter < n_estimators; ++iter) {
+            // Train base estimator with current sample weights
+            auto estimator = trainBaseEstimator(sample_weights);
+
+            // Calculate weighted error
+            double weighted_error = calculateWeightedError(estimator.get(), sample_weights);
+            training_errors.push_back(weighted_error);
+
+            // According to SAMME, we need error < random_guess_error
+            if (weighted_error >= random_guess_error) {
+                DEBUG_LOG(debug, "Error >= random guess (" << random_guess_error << "), stopping");
+                // If only one estimator and it's worse than random, keep it with zero weight
+                if (models.empty()) {
+                    models.push_back(std::move(estimator));
+                    alphas.push_back(0.0);
+                }
+                break;  // Stop boosting
+            }
+
+            // Check for perfect classification BEFORE calculating alpha
+            if (weighted_error <= 1e-10) {
+                DEBUG_LOG(debug, "Perfect classification achieved (error=" << weighted_error << ")");
+
+                // For perfect classification, use a large but finite alpha
+                double alpha = 10.0 + std::log(static_cast<double>(n_classes - 1));
+
+                // Store the estimator and its weight
+                models.push_back(std::move(estimator));
+                alphas.push_back(alpha);
+
+                DEBUG_LOG(debug, "Iteration " << iter << ":\n"
+                    << "  Weighted error: " << weighted_error << "\n"
+                    << "  Alpha (finite): " << alpha << "\n"
+                    << "  Random guess error: " << random_guess_error);
+
+                break;  // Stop training as we have a perfect classifier
+            }
+
+            // Calculate alpha (estimator weight) using SAMME formula
+            // alpha = log((1 - err) / err) + log(K - 1)
+            // Clamp weighted_error to avoid division by zero and infinite alpha
+            double clamped_error = std::max(1e-15, std::min(1.0 - 1e-15, weighted_error));
+            double alpha = std::log((1.0 - clamped_error) / clamped_error) +
+                std::log(static_cast<double>(n_classes - 1));
+
+            // Clamp alpha to reasonable bounds to avoid numerical issues
+            alpha = std::max(-10.0, std::min(10.0, alpha));
+
+            // Store the estimator and its weight
+            models.push_back(std::move(estimator));
+            alphas.push_back(alpha);
+
+            // Update sample weights (only if this is not the last iteration)
+            if (iter < n_estimators - 1) {
+                updateSampleWeights(models.back().get(), alpha);
+                normalizeWeights();
+            }
+
+            DEBUG_LOG(debug, "Iteration " << iter << ":\n"
+                << "  Weighted error: " << weighted_error << "\n"
+                << "  Alpha: " << alpha << "\n"
+                << "  Random guess error: " << random_guess_error);
+        }
+
+        // Set the number of models actually trained
+        n_models = models.size();
+        DEBUG_LOG(debug, "AdaBoost training completed with " << n_models << " models");
+    }
+
+    void AdaBoost::trainModel(const torch::Tensor& weights, const Smoothing_t smoothing)
+    {
+        // Call buildModel which does the actual training
+        buildModel(weights);
+        fitted = true;
+    }
+
+    std::unique_ptr<Classifier> AdaBoost::trainBaseEstimator(const torch::Tensor& weights)
+    {
+        // Create a decision tree with specified max depth
+        auto tree = std::make_unique<DecisionTree>(base_max_depth);
+
+        // Ensure weights are properly normalized
+        auto normalized_weights = weights / weights.sum();
+
+        // Fit the tree with the current sample weights
+        tree->fit(dataset, features, className, states, normalized_weights, Smoothing_t::NONE);
+
+        return tree;
+    }
+
+    double AdaBoost::calculateWeightedError(Classifier* estimator, const torch::Tensor& weights)
+    {
+        // Get features and labels from dataset (avoid repeated indexing)
+        auto X = dataset.index({ torch::indexing::Slice(0, dataset.size(0) - 1), torch::indexing::Slice() });
+        auto y_true = dataset.index({ -1, torch::indexing::Slice() });
+
+        // Get predictions from the estimator
+        auto y_pred = estimator->predict(X);
+
+        // Vectorized error calculation using PyTorch operations
+        auto incorrect = (y_pred != y_true).to(torch::kDouble);
+
+        // Direct dot product for weighted error (more efficient than sum)
+        double weighted_error = torch::dot(incorrect, weights).item<double>();
+
+        // Clamp to valid range in one operation
+        return std::clamp(weighted_error, 1e-15, 1.0 - 1e-15);
+    }
+
+    void AdaBoost::updateSampleWeights(Classifier* estimator, double alpha)
+    {
+        // Get predictions from the estimator (reuse from calculateWeightedError if possible)
+        auto X = dataset.index({ torch::indexing::Slice(0, dataset.size(0) - 1), torch::indexing::Slice() });
+        auto y_true = dataset.index({ -1, torch::indexing::Slice() });
+        auto y_pred = estimator->predict(X);
+
+        // Vectorized weight update using PyTorch operations
+        auto incorrect = (y_pred != y_true).to(torch::kDouble);
+
+        // Single vectorized operation instead of element-wise multiplication
+        sample_weights *= torch::exp(alpha * incorrect);
+
+        // Vectorized clamping for numerical stability
+        sample_weights = torch::clamp(sample_weights, 1e-15, 1e15);
+    }
+
+    void AdaBoost::normalizeWeights()
+    {
+        // Single-pass normalization using PyTorch operations
+        double sum_weights = torch::sum(sample_weights).item<double>();
+
+        if (__builtin_expect(sum_weights <= 0, 0)) {
+            // Reset to uniform if all weights are zero/negative (rare case)
+            sample_weights = torch::ones_like(sample_weights) / sample_weights.size(0);
+        } else {
+            // Vectorized normalization
+            sample_weights /= sum_weights;
+
+            // Vectorized minimum weight enforcement
+            sample_weights = torch::clamp_min(sample_weights, 1e-15);
+
+            // Renormalize after clamping (if any weights were clamped)
+            double new_sum = torch::sum(sample_weights).item<double>();
+            if (new_sum != 1.0) {
+                sample_weights /= new_sum;
+            }
+        }
+    }
+
+    std::vector<std::string> AdaBoost::graph(const std::string& title) const
+    {
+        // Create a graph representation of the AdaBoost ensemble
+        std::vector<std::string> graph_lines;
+
+        // Header
+        graph_lines.push_back("digraph AdaBoost {");
+        graph_lines.push_back("    rankdir=TB;");
+        graph_lines.push_back("    node [shape=box];");
+
+        if (!title.empty()) {
+            graph_lines.push_back("    label=\"" + title + "\";");
+            graph_lines.push_back("    labelloc=t;");
+        }
+
+        // Add input node
+        graph_lines.push_back("    Input [shape=ellipse, label=\"Input Features\"];");
+
+        // Add base estimators
+        for (size_t i = 0; i < models.size(); ++i) {
+            std::stringstream ss;
+            ss << "    Estimator" << i << " [label=\"Base Estimator " << i + 1
+                << "\\nα = " << std::fixed << std::setprecision(3) << alphas[i] << "\"];";
+            graph_lines.push_back(ss.str());
+
+            // Connect input to estimator
+            ss.str("");
+            ss << "    Input -> Estimator" << i << ";";
+            graph_lines.push_back(ss.str());
+        }
+
+        // Add combination node
+        graph_lines.push_back("    Combination [shape=diamond, label=\"Weighted Vote\"];");
+
+        // Connect estimators to combination
+        for (size_t i = 0; i < models.size(); ++i) {
+            std::stringstream ss;
+            ss << "    Estimator" << i << " -> Combination;";
+            graph_lines.push_back(ss.str());
+        }
+
+        // Add output node
+        graph_lines.push_back("    Output [shape=ellipse, label=\"Final Prediction\"];");
+        graph_lines.push_back("    Combination -> Output;");
+
+        // Close graph
+        graph_lines.push_back("}");
+
+        return graph_lines;
+    }
+
+    void AdaBoost::checkValues() const
+    {
+        if (n_estimators <= 0) {
+            throw std::invalid_argument("n_estimators must be positive");
+        }
+        if (base_max_depth <= 0) {
+            throw std::invalid_argument("base_max_depth must be positive");
+        }
+    }
+
+    void AdaBoost::setHyperparameters(const nlohmann::json& hyperparameters_)
+    {
+        auto hyperparameters = hyperparameters_;
+        // Set hyperparameters from JSON
+        auto it = hyperparameters.find("n_estimators");
+        if (it != hyperparameters.end()) {
+            n_estimators = it->get<int>();
+            hyperparameters.erase("n_estimators");
+        }
+
+        it = hyperparameters.find("base_max_depth");
+        if (it != hyperparameters.end()) {
+            base_max_depth = it->get<int>();
+            hyperparameters.erase("base_max_depth");
+        }
+        checkValues();
+        Ensemble::setHyperparameters(hyperparameters);
+    }
+
+    int AdaBoost::predictSample(const torch::Tensor& x) const
+    {
+        // Early validation (keep essential checks only)
+        if (!fitted || models.empty()) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        // Pre-allocate and reuse memory
+        static thread_local std::vector<double> class_votes_cache;
+        if (class_votes_cache.size() != static_cast<size_t>(n_classes)) {
+            class_votes_cache.resize(n_classes);
+        }
+        std::fill(class_votes_cache.begin(), class_votes_cache.end(), 0.0);
+
+        // Optimized voting loop - avoid exception handling in hot path
+        for (size_t i = 0; i < models.size(); ++i) {
+            double alpha = alphas[i];
+            if (alpha <= 0 || !std::isfinite(alpha)) continue;
+
+            // Direct cast and call - avoid virtual dispatch overhead
+            int predicted_class = static_cast<DecisionTree*>(models[i].get())->predictSample(x);
+
+            // Bounds check with branch prediction hint
+            if (__builtin_expect(predicted_class >= 0 && predicted_class < n_classes, 1)) {
+                class_votes_cache[predicted_class] += alpha;
+            }
+        }
+
+        // Fast argmax using iterators
+        return std::distance(class_votes_cache.begin(),
+            std::max_element(class_votes_cache.begin(), class_votes_cache.end()));
+    }
+
+    torch::Tensor AdaBoost::predictProbaSample(const torch::Tensor& x) const
+    {
+        // Early validation
+        if (!fitted || models.empty()) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        // Use stack allocation for small arrays (typical case: n_classes <= 32)
+        constexpr int STACK_THRESHOLD = 32;
+        double stack_votes[STACK_THRESHOLD];
+        std::vector<double> heap_votes;
+        double* class_votes;
+
+        if (n_classes <= STACK_THRESHOLD) {
+            class_votes = stack_votes;
+            std::fill_n(class_votes, n_classes, 0.0);
+        } else {
+            heap_votes.resize(n_classes, 0.0);
+            class_votes = heap_votes.data();
+        }
+
+        double total_votes = 0.0;
+
+        // Optimized voting loop
+        for (size_t i = 0; i < models.size(); ++i) {
+            double alpha = alphas[i];
+            if (alpha <= 0 || !std::isfinite(alpha)) continue;
+
+            int predicted_class = static_cast<DecisionTree*>(models[i].get())->predictSample(x);
+
+            if (__builtin_expect(predicted_class >= 0 && predicted_class < n_classes, 1)) {
+                class_votes[predicted_class] += alpha;
+                total_votes += alpha;
+            }
+        }
+
+        // Direct tensor creation with pre-computed size
+        torch::Tensor class_probs = torch::empty({ n_classes }, torch::TensorOptions().dtype(torch::kFloat32));
+        auto probs_accessor = class_probs.accessor<float, 1>();
+
+        if (__builtin_expect(total_votes > 0.0, 1)) {
+            // Vectorized probability calculation
+            const double inv_total = 1.0 / total_votes;
+            for (int j = 0; j < n_classes; ++j) {
+                probs_accessor[j] = static_cast<float>(class_votes[j] * inv_total);
+            }
+        } else {
+            // Uniform distribution fallback
+            const float uniform_prob = 1.0f / n_classes;
+            for (int j = 0; j < n_classes; ++j) {
+                probs_accessor[j] = uniform_prob;
+            }
+        }
+
+        return class_probs;
+    }
+
+    torch::Tensor AdaBoost::predict_proba(torch::Tensor& X)
+    {
+        if (!fitted || models.empty()) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        // Input validation
+        if (X.size(0) != n) {
+            throw std::runtime_error("Input has wrong number of features. Expected " +
+                std::to_string(n) + " but got " + std::to_string(X.size(0)));
+        }
+
+        const int n_samples = X.size(1);
+
+        // Pre-allocate output tensor with correct layout
+        torch::Tensor probabilities = torch::empty({ n_samples, n_classes },
+            torch::TensorOptions().dtype(torch::kFloat32));
+
+        // Convert to contiguous memory if needed (optimization for memory access)
+        if (!X.is_contiguous()) {
+            X = X.contiguous();
+        }
+
+        // Batch processing with memory-efficient sample extraction
+        for (int i = 0; i < n_samples; ++i) {
+            // Extract sample without unnecessary copies
+            auto sample = X.select(1, i);
+
+            // Direct assignment to pre-allocated tensor
+            probabilities[i] = predictProbaSample(sample);
+        }
+
+        return probabilities;
+    }
+
+    std::vector<std::vector<double>> AdaBoost::predict_proba(std::vector<std::vector<int>>& X)
+    {
+        const size_t n_samples = X[0].size();
+
+        // Pre-allocate result with exact size
+        std::vector<std::vector<double>> result;
+        result.reserve(n_samples);
+
+        // Avoid repeated allocations
+        for (size_t i = 0; i < n_samples; ++i) {
+            result.emplace_back(n_classes, 0.0);
+        }
+
+        // Convert to tensor only once (batch conversion is more efficient)
+        torch::Tensor X_tensor = platform::TensorUtils::to_matrix(X);
+        torch::Tensor proba_tensor = predict_proba(X_tensor);
+
+        // Optimized tensor-to-vector conversion
+        auto proba_accessor = proba_tensor.accessor<float, 2>();
+        for (size_t i = 0; i < n_samples; ++i) {
+            for (int j = 0; j < n_classes; ++j) {
+                result[i][j] = static_cast<double>(proba_accessor[i][j]);
+            }
+        }
+
+        return result;
+    }
+
+    torch::Tensor AdaBoost::predict(torch::Tensor& X)
+    {
+        if (!fitted || models.empty()) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        if (X.size(0) != n) {
+            throw std::runtime_error("Input has wrong number of features. Expected " +
+                std::to_string(n) + " but got " + std::to_string(X.size(0)));
+        }
+
+        const int n_samples = X.size(1);
+
+        // Pre-allocate with correct dtype
+        torch::Tensor predictions = torch::empty({ n_samples }, torch::TensorOptions().dtype(torch::kInt32));
+        auto pred_accessor = predictions.accessor<int32_t, 1>();
+
+        // Ensure contiguous memory layout
+        if (!X.is_contiguous()) {
+            X = X.contiguous();
+        }
+
+        // Optimized prediction loop
+        for (int i = 0; i < n_samples; ++i) {
+            auto sample = X.select(1, i);
+            pred_accessor[i] = predictSample(sample);
+        }
+
+        return predictions;
+    }
+
+    std::vector<int> AdaBoost::predict(std::vector<std::vector<int>>& X)
+    {
+        // Single tensor conversion for batch processing
+        torch::Tensor X_tensor = platform::TensorUtils::to_matrix(X);
+        torch::Tensor predictions_tensor = predict(X_tensor);
+
+        // Optimized tensor-to-vector conversion
+        std::vector<int> result = platform::TensorUtils::to_vector<int>(predictions_tensor);
+        return result;
+    }
+
+} // namespace bayesnet

src/experimental_clfs/AdaBoost.h

@@ -0,0 +1,81 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#ifndef ADABOOST_H
+#define ADABOOST_H
+
+#include <vector>
+#include <memory>
+#include "bayesnet/ensembles/Ensemble.h"
+
+namespace bayesnet {
+    class AdaBoost : public Ensemble {
+    public:
+        explicit AdaBoost(int n_estimators = 100, int max_depth = 1);
+        virtual ~AdaBoost() = default;
+
+        // Override base class methods
+        std::vector<std::string> graph(const std::string& title = "") const override;
+
+        // AdaBoost specific methods
+        void setNEstimators(int n_estimators) { this->n_estimators = n_estimators; checkValues(); }
+        int getNEstimators() const { return n_estimators; }
+        void setBaseMaxDepth(int depth) { this->base_max_depth = depth; checkValues(); }
+        int getBaseMaxDepth() const { return base_max_depth; }
+
+        // Get the weight of each base estimator
+        std::vector<double> getEstimatorWeights() const { return alphas; }
+
+        // Get training errors for each iteration
+        std::vector<double> getTrainingErrors() const { return training_errors; }
+
+        // Override setHyperparameters from BaseClassifier
+        void setHyperparameters(const nlohmann::json& hyperparameters) override;
+
+        torch::Tensor predict(torch::Tensor& X) override;
+        std::vector<int> predict(std::vector<std::vector<int>>& X) override;
+        torch::Tensor predict_proba(torch::Tensor& X) override;
+        std::vector<std::vector<double>> predict_proba(std::vector<std::vector<int>>& X);
+        void setDebug(bool debug) { this->debug = debug; }
+
+    protected:
+        void buildModel(const torch::Tensor& weights) override;
+        void trainModel(const torch::Tensor& weights, const Smoothing_t smoothing) override;
+
+    private:
+        int n_estimators;
+        int base_max_depth;  // Max depth for base decision trees
+        std::vector<double> alphas;  // Weight of each base estimator
+        std::vector<double> training_errors;  // Training error at each iteration
+        torch::Tensor sample_weights;  // Current sample weights
+        int n_classes;  // Number of classes in the target variable
+        int n;  // Number of features
+
+        // Train a single base estimator
+        std::unique_ptr<Classifier> trainBaseEstimator(const torch::Tensor& weights);
+
+        // Calculate weighted error
+        double calculateWeightedError(Classifier* estimator, const torch::Tensor& weights);
+
+        // Update sample weights based on predictions
+        void updateSampleWeights(Classifier* estimator, double alpha);
+
+        // Normalize weights to sum to 1
+        void normalizeWeights();
+
+        // Check if hyperparameters values are valid
+        void checkValues() const;
+
+        // Make predictions for a single sample
+        int predictSample(const torch::Tensor& x) const;
+
+        // Make probabilistic predictions for a single sample
+        torch::Tensor predictProbaSample(const torch::Tensor& x) const;
+        bool debug = false;  // Enable debug mode for debug output
+    };
+}
+
+#endif // ADABOOST_H

src/experimental_clfs/CountingSemaphore.hpp

@@ -0,0 +1,53 @@
+#ifndef COUNTING_SEMAPHORE_H
+#define COUNTING_SEMAPHORE_H
+#include <mutex>
+#include <condition_variable>
+#include <algorithm>
+#include <thread>
+#include <mutex>
+#include <condition_variable>
+
+class CountingSemaphore {
+public:
+    static CountingSemaphore& getInstance()
+    {
+        static CountingSemaphore instance;
+        return instance;
+    }
+    // Delete copy constructor and assignment operator
+    CountingSemaphore(const CountingSemaphore&) = delete;
+    CountingSemaphore& operator=(const CountingSemaphore&) = delete;
+    void acquire()
+    {
+        std::unique_lock<std::mutex> lock(mtx_);
+        cv_.wait(lock, [this]() { return count_ > 0; });
+        --count_;
+    }
+    void release()
+    {
+        std::lock_guard<std::mutex> lock(mtx_);
+        ++count_;
+        if (count_ <= max_count_) {
+            cv_.notify_one();
+        }
+    }
+    uint getCount() const
+    {
+        return count_;
+    }
+    uint getMaxCount() const
+    {
+        return max_count_;
+    }
+private:
+    CountingSemaphore()
+        : max_count_(std::max(1u, static_cast<uint>(0.95 * std::thread::hardware_concurrency()))),
+        count_(max_count_)
+    {
+    }
+    std::mutex mtx_;
+    std::condition_variable cv_;
+    const uint max_count_;
+    uint count_;
+};
+#endif

src/experimental_clfs/DecisionTree.cpp

@@ -0,0 +1,495 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include "DecisionTree.h"
+#include <algorithm>
+#include <numeric>
+#include <sstream>
+#include <iomanip>
+#include <limits>
+#include "TensorUtils.hpp"
+
+namespace bayesnet {
+
+    DecisionTree::DecisionTree(int max_depth, int min_samples_split, int min_samples_leaf)
+        : Classifier(Network()), max_depth(max_depth),
+        min_samples_split(min_samples_split), min_samples_leaf(min_samples_leaf)
+    {
+        validHyperparameters = { "max_depth", "min_samples_split", "min_samples_leaf" };
+    }
+
+    void DecisionTree::setHyperparameters(const nlohmann::json& hyperparameters_)
+    {
+        auto hyperparameters = hyperparameters_;
+        // Set hyperparameters from JSON
+        auto it = hyperparameters.find("max_depth");
+        if (it != hyperparameters.end()) {
+            max_depth = it->get<int>();
+            hyperparameters.erase("max_depth");  // Remove 'order' if present
+        }
+
+        it = hyperparameters.find("min_samples_split");
+        if (it != hyperparameters.end()) {
+            min_samples_split = it->get<int>();
+            hyperparameters.erase("min_samples_split");  // Remove 'min_samples_split' if present   
+        }
+
+        it = hyperparameters.find("min_samples_leaf");
+        if (it != hyperparameters.end()) {
+            min_samples_leaf = it->get<int>();
+            hyperparameters.erase("min_samples_leaf");  // Remove 'min_samples_leaf' if present
+        }
+        Classifier::setHyperparameters(hyperparameters);
+        checkValues();
+    }
+    void DecisionTree::checkValues()
+    {
+        if (max_depth <= 0) {
+            throw std::invalid_argument("max_depth must be positive");
+        }
+        if (min_samples_leaf <= 0) {
+            throw std::invalid_argument("min_samples_leaf must be positive");
+        }
+        if (min_samples_split <= 0) {
+            throw std::invalid_argument("min_samples_split must be positive");
+        }
+    }
+    void DecisionTree::buildModel(const torch::Tensor& weights)
+    {
+        // Extract features (X) and labels (y) from dataset
+        auto X = dataset.index({ torch::indexing::Slice(0, dataset.size(0) - 1), torch::indexing::Slice() }).t();
+        auto y = dataset.index({ -1, torch::indexing::Slice() });
+
+        if (X.size(0) != y.size(0)) {
+            throw std::runtime_error("X and y must have the same number of samples");
+        }
+
+        n_classes = states[className].size();
+
+        // Use provided weights or uniform weights
+        torch::Tensor sample_weights;
+        if (weights.defined() && weights.numel() > 0) {
+            if (weights.size(0) != X.size(0)) {
+                throw std::runtime_error("weights must have the same length as number of samples");
+            }
+            sample_weights = weights;
+        } else {
+            sample_weights = torch::ones({ X.size(0) }) / X.size(0);
+        }
+
+        // Normalize weights
+        sample_weights = sample_weights / sample_weights.sum();
+
+        // Build the tree
+        root = buildTree(X, y, sample_weights, 0);
+
+        // Mark as fitted
+        fitted = true;
+    }
+    bool DecisionTree::validateTensors(const torch::Tensor& X, const torch::Tensor& y,
+        const torch::Tensor& sample_weights) const
+    {
+        if (X.size(0) != y.size(0) || X.size(0) != sample_weights.size(0)) {
+            return false;
+        }
+        if (X.size(0) == 0) {
+            return false;
+        }
+        return true;
+    }
+
+    std::unique_ptr<TreeNode> DecisionTree::buildTree(
+        const torch::Tensor& X,
+        const torch::Tensor& y,
+        const torch::Tensor& sample_weights,
+        int current_depth)
+    {
+        auto node = std::make_unique<TreeNode>();
+        int n_samples = y.size(0);
+
+        // Check stopping criteria
+        auto unique = at::_unique(y);
+        bool should_stop = (current_depth >= max_depth) ||
+            (n_samples < min_samples_split) ||
+            (std::get<0>(unique).size(0) == 1);  // All samples same class
+
+        if (should_stop || n_samples <= min_samples_leaf) {
+            // Create leaf node
+            node->is_leaf = true;
+
+            // Calculate class probabilities
+            node->class_probabilities = torch::zeros({ n_classes });
+
+            for (int i = 0; i < n_samples; i++) {
+                int class_idx = y[i].item<int>();
+                node->class_probabilities[class_idx] += sample_weights[i].item<float>();
+            }
+
+            // Normalize probabilities
+            node->class_probabilities /= node->class_probabilities.sum();
+
+            // Set predicted class as the one with highest probability
+            node->predicted_class = torch::argmax(node->class_probabilities).item<int>();
+
+            return node;
+        }
+
+        // Find best split
+        SplitInfo best_split = findBestSplit(X, y, sample_weights);
+
+        // If no valid split found, create leaf
+        if (best_split.feature_index == -1 || best_split.impurity_decrease <= 0) {
+            node->is_leaf = true;
+
+            // Calculate class probabilities
+            node->class_probabilities = torch::zeros({ n_classes });
+
+            for (int i = 0; i < n_samples; i++) {
+                int class_idx = y[i].item<int>();
+                node->class_probabilities[class_idx] += sample_weights[i].item<float>();
+            }
+
+            node->class_probabilities /= node->class_probabilities.sum();
+            node->predicted_class = torch::argmax(node->class_probabilities).item<int>();
+
+            return node;
+        }
+
+        // Create internal node
+        node->is_leaf = false;
+        node->split_feature = best_split.feature_index;
+        node->split_value = best_split.split_value;
+
+        // Split data
+        auto left_X = X.index({ best_split.left_mask });
+        auto left_y = y.index({ best_split.left_mask });
+        auto left_weights = sample_weights.index({ best_split.left_mask });
+
+        auto right_X = X.index({ best_split.right_mask });
+        auto right_y = y.index({ best_split.right_mask });
+        auto right_weights = sample_weights.index({ best_split.right_mask });
+
+        // Recursively build subtrees
+        if (left_X.size(0) >= min_samples_leaf) {
+            node->left = buildTree(left_X, left_y, left_weights, current_depth + 1);
+        } else {
+            // Force leaf if not enough samples
+            node->left = std::make_unique<TreeNode>();
+            node->left->is_leaf = true;
+            auto mode = std::get<0>(torch::mode(left_y));
+            node->left->predicted_class = mode.item<int>();
+            node->left->class_probabilities = torch::zeros({ n_classes });
+            node->left->class_probabilities[node->left->predicted_class] = 1.0;
+        }
+
+        if (right_X.size(0) >= min_samples_leaf) {
+            node->right = buildTree(right_X, right_y, right_weights, current_depth + 1);
+        } else {
+            // Force leaf if not enough samples
+            node->right = std::make_unique<TreeNode>();
+            node->right->is_leaf = true;
+            auto mode = std::get<0>(torch::mode(right_y));
+            node->right->predicted_class = mode.item<int>();
+            node->right->class_probabilities = torch::zeros({ n_classes });
+            node->right->class_probabilities[node->right->predicted_class] = 1.0;
+        }
+
+        return node;
+    }
+
+    DecisionTree::SplitInfo DecisionTree::findBestSplit(
+        const torch::Tensor& X,
+        const torch::Tensor& y,
+        const torch::Tensor& sample_weights)
+    {
+
+        SplitInfo best_split;
+        best_split.feature_index = -1;
+        best_split.split_value = -1;
+        best_split.impurity_decrease = -std::numeric_limits<double>::infinity();
+
+        int n_features = X.size(1);
+        int n_samples = X.size(0);
+
+        // Calculate impurity of current node
+        double current_impurity = calculateGiniImpurity(y, sample_weights);
+        double total_weight = sample_weights.sum().item<double>();
+
+        // Try each feature
+        for (int feat_idx = 0; feat_idx < n_features; feat_idx++) {
+            auto feature_values = X.index({ torch::indexing::Slice(), feat_idx });
+            auto unique_values = std::get<0>(torch::unique_consecutive(std::get<0>(torch::sort(feature_values))));
+
+            // Try each unique value as split point
+            for (int i = 0; i < unique_values.size(0); i++) {
+                int split_val = unique_values[i].item<int>();
+
+                // Create masks for left and right splits
+                auto left_mask = feature_values == split_val;
+                auto right_mask = ~left_mask;
+
+                int left_count = left_mask.sum().item<int>();
+                int right_count = right_mask.sum().item<int>();
+
+                // Skip if split doesn't satisfy minimum samples requirement
+                if (left_count < min_samples_leaf || right_count < min_samples_leaf) {
+                    continue;
+                }
+
+                // Calculate weighted impurities
+                auto left_y = y.index({ left_mask });
+                auto left_weights = sample_weights.index({ left_mask });
+                double left_weight = left_weights.sum().item<double>();
+                double left_impurity = calculateGiniImpurity(left_y, left_weights);
+
+                auto right_y = y.index({ right_mask });
+                auto right_weights = sample_weights.index({ right_mask });
+                double right_weight = right_weights.sum().item<double>();
+                double right_impurity = calculateGiniImpurity(right_y, right_weights);
+
+                // Calculate impurity decrease
+                double impurity_decrease = current_impurity -
+                    (left_weight / total_weight * left_impurity +
+                        right_weight / total_weight * right_impurity);
+
+                // Update best split if this is better
+                if (impurity_decrease > best_split.impurity_decrease) {
+                    best_split.feature_index = feat_idx;
+                    best_split.split_value = split_val;
+                    best_split.impurity_decrease = impurity_decrease;
+                    best_split.left_mask = left_mask;
+                    best_split.right_mask = right_mask;
+                }
+            }
+        }
+
+        return best_split;
+    }
+
+    double DecisionTree::calculateGiniImpurity(
+        const torch::Tensor& y,
+        const torch::Tensor& sample_weights)
+    {
+        if (y.size(0) == 0 || sample_weights.size(0) == 0) {
+            return 0.0;
+        }
+
+        if (y.size(0) != sample_weights.size(0)) {
+            throw std::runtime_error("y and sample_weights must have same size");
+        }
+
+        torch::Tensor class_weights = torch::zeros({ n_classes });
+
+        // Calculate weighted class counts
+        for (int i = 0; i < y.size(0); i++) {
+            int class_idx = y[i].item<int>();
+
+            if (class_idx < 0 || class_idx >= n_classes) {
+                throw std::runtime_error("Invalid class index: " + std::to_string(class_idx));
+            }
+
+            class_weights[class_idx] += sample_weights[i].item<float>();
+        }
+
+        // Normalize
+        double total_weight = class_weights.sum().item<double>();
+        if (total_weight == 0) return 0.0;
+
+        class_weights /= total_weight;
+
+        // Calculate Gini impurity: 1 - sum(p_i^2)
+        double gini = 1.0;
+        for (int i = 0; i < n_classes; i++) {
+            double p = class_weights[i].item<double>();
+            gini -= p * p;
+        }
+
+        return gini;
+    }
+
+
+    torch::Tensor DecisionTree::predict(torch::Tensor& X)
+    {
+        if (!fitted) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        int n_samples = X.size(1);
+        torch::Tensor predictions = torch::zeros({ n_samples }, torch::kInt32);
+
+        for (int i = 0; i < n_samples; i++) {
+            auto sample = X.index({ torch::indexing::Slice(), i }).ravel();
+            predictions[i] = predictSample(sample);
+        }
+
+        return predictions;
+    }
+
+    std::vector<int> DecisionTree::predict(std::vector<std::vector<int>>& X)
+    {
+        // Convert to tensor
+        long n = X.size();
+        long m = X.at(0).size();
+        torch::Tensor X_tensor = platform::TensorUtils::to_matrix(X);
+        auto predictions = predict(X_tensor);
+        std::vector<int> result = platform::TensorUtils::to_vector<int>(predictions);
+        return result;
+    }
+
+    torch::Tensor DecisionTree::predict_proba(torch::Tensor& X)
+    {
+        if (!fitted) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        int n_samples = X.size(1);
+        torch::Tensor probabilities = torch::zeros({ n_samples, n_classes });
+
+        for (int i = 0; i < n_samples; i++) {
+            auto sample = X.index({ torch::indexing::Slice(), i }).ravel();
+            probabilities[i] = predictProbaSample(sample);
+        }
+
+        return probabilities;
+    }
+
+    std::vector<std::vector<double>> DecisionTree::predict_proba(std::vector<std::vector<int>>& X)
+    {
+        auto n_samples = X.at(0).size();
+        // Convert to tensor
+        torch::Tensor X_tensor = platform::TensorUtils::to_matrix(X);
+        auto proba_tensor = predict_proba(X_tensor);
+        std::vector<std::vector<double>> result(n_samples, std::vector<double>(n_classes, 0.0));
+
+        for (int i = 0; i < n_samples; i++) {
+            for (int j = 0; j < n_classes; j++) {
+                result[i][j] = proba_tensor[i][j].item<double>();
+            }
+        }
+
+        return result;
+    }
+
+    int DecisionTree::predictSample(const torch::Tensor& x) const
+    {
+        if (!fitted) {
+            throw std::runtime_error(CLASSIFIER_NOT_FITTED);
+        }
+
+        if (x.size(0) != n) {  // n debería ser el número de características
+            throw std::runtime_error("Input sample has wrong number of features");
+        }
+
+        const TreeNode* leaf = traverseTree(x, root.get());
+        return leaf->predicted_class;
+    }
+    torch::Tensor DecisionTree::predictProbaSample(const torch::Tensor& x) const
+    {
+        const TreeNode* leaf = traverseTree(x, root.get());
+        return leaf->class_probabilities.clone();
+    }
+
+
+    const TreeNode* DecisionTree::traverseTree(const torch::Tensor& x, const TreeNode* node) const
+    {
+        if (!node) {
+            throw std::runtime_error("Null node encountered during tree traversal");
+        }
+
+        if (node->is_leaf) {
+            return node;
+        }
+
+        if (node->split_feature < 0 || node->split_feature >= x.size(0)) {
+            throw std::runtime_error("Invalid split_feature index: " + std::to_string(node->split_feature));
+        }
+
+        int feature_value = x[node->split_feature].item<int>();
+
+        if (feature_value == node->split_value) {
+            if (!node->left) {
+                throw std::runtime_error("Missing left child in tree");
+            }
+            return traverseTree(x, node->left.get());
+        } else {
+            if (!node->right) {
+                throw std::runtime_error("Missing right child in tree");
+            }
+            return traverseTree(x, node->right.get());
+        }
+    }
+
+    std::vector<std::string> DecisionTree::graph(const std::string& title) const
+    {
+        std::vector<std::string> lines;
+        lines.push_back("digraph DecisionTree {");
+        lines.push_back("    rankdir=TB;");
+        lines.push_back("    node [shape=box, style=\"filled, rounded\", fontname=\"helvetica\"];");
+        lines.push_back("    edge [fontname=\"helvetica\"];");
+
+        if (!title.empty()) {
+            lines.push_back("    label=\"" + title + "\";");
+            lines.push_back("    labelloc=t;");
+        }
+
+        if (root) {
+            int node_id = 0;
+            treeToGraph(root.get(), lines, node_id);
+        }
+
+        lines.push_back("}");
+        return lines;
+    }
+
+    void DecisionTree::treeToGraph(
+        const TreeNode* node,
+        std::vector<std::string>& lines,
+        int& node_id,
+        int parent_id,
+        const std::string& edge_label) const
+    {
+
+        int current_id = node_id++;
+        std::stringstream ss;
+
+        if (node->is_leaf) {
+            // Leaf node
+            ss << "    node" << current_id << " [label=\"Class: " << node->predicted_class;
+            ss << "\\nProb: " << std::fixed << std::setprecision(3)
+                << node->class_probabilities[node->predicted_class].item<float>();
+            ss << "\", fillcolor=\"lightblue\"];";
+            lines.push_back(ss.str());
+        } else {
+            // Internal node
+            ss << "    node" << current_id << " [label=\"" << features[node->split_feature];
+            ss << " = " << node->split_value << "?\", fillcolor=\"lightgreen\"];";
+            lines.push_back(ss.str());
+        }
+
+        // Add edge from parent
+        if (parent_id >= 0) {
+            ss.str("");
+            ss << "    node" << parent_id << " -> node" << current_id;
+            if (!edge_label.empty()) {
+                ss << " [label=\"" << edge_label << "\"];";
+            } else {
+                ss << ";";
+            }
+            lines.push_back(ss.str());
+        }
+
+        // Recurse on children
+        if (!node->is_leaf) {
+            if (node->left) {
+                treeToGraph(node->left.get(), lines, node_id, current_id, "Yes");
+            }
+            if (node->right) {
+                treeToGraph(node->right.get(), lines, node_id, current_id, "No");
+            }
+        }
+    }
+
+} // namespace bayesnet

src/experimental_clfs/DecisionTree.h

@@ -0,0 +1,134 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#ifndef DECISION_TREE_H
+#define DECISION_TREE_H
+
+#include <memory>
+#include <vector>
+#include <map>
+#include <torch/torch.h>
+#include "bayesnet/classifiers/Classifier.h"
+
+namespace bayesnet {
+
+    // Forward declaration
+    struct TreeNode;
+
+    class DecisionTree : public Classifier {
+    public:
+        explicit DecisionTree(int max_depth = 3, int min_samples_split = 2, int min_samples_leaf = 1);
+        virtual ~DecisionTree() = default;
+
+        // Override graph method to show tree structure
+        std::vector<std::string> graph(const std::string& title = "") const override;
+
+        // Setters for hyperparameters
+        void setMaxDepth(int depth) { max_depth = depth; checkValues(); }
+        void setMinSamplesSplit(int samples) { min_samples_split = samples; checkValues(); }
+        void setMinSamplesLeaf(int samples) { min_samples_leaf = samples; checkValues(); }
+        int getMaxDepth() const { return max_depth; }
+        int getMinSamplesSplit() const { return min_samples_split; }
+        int getMinSamplesLeaf() const { return min_samples_leaf; }
+
+        // Override setHyperparameters
+        void setHyperparameters(const nlohmann::json& hyperparameters) override;
+
+        torch::Tensor predict(torch::Tensor& X) override;
+        std::vector<int> predict(std::vector<std::vector<int>>& X) override;
+        torch::Tensor predict_proba(torch::Tensor& X) override;
+        std::vector<std::vector<double>> predict_proba(std::vector<std::vector<int>>& X);
+
+        // Make predictions for a single sample
+        int predictSample(const torch::Tensor& x) const;
+
+        // Make probabilistic predictions for a single sample
+        torch::Tensor predictProbaSample(const torch::Tensor& x) const;
+
+    protected:
+        void buildModel(const torch::Tensor& weights) override;
+        void trainModel(const torch::Tensor& weights, const Smoothing_t smoothing) override
+        {
+            // Decision trees do not require training in the traditional sense
+            // as they are built from the data directly.
+            // This method can be used to set weights or other parameters if needed.
+        }
+    private:
+        void checkValues();
+        bool validateTensors(const torch::Tensor& X, const torch::Tensor& y, const torch::Tensor& sample_weights) const;
+        // Tree hyperparameters
+        int max_depth;
+        int min_samples_split;
+        int min_samples_leaf;
+        int n_classes;  // Number of classes in the target variable
+
+        // Root of the decision tree
+        std::unique_ptr<TreeNode> root;
+
+        // Build tree recursively
+        std::unique_ptr<TreeNode> buildTree(
+            const torch::Tensor& X,
+            const torch::Tensor& y,
+            const torch::Tensor& sample_weights,
+            int current_depth
+        );
+
+        // Find best split for a node
+        struct SplitInfo {
+            int feature_index;
+            int split_value;
+            double impurity_decrease;
+            torch::Tensor left_mask;
+            torch::Tensor right_mask;
+        };
+
+        SplitInfo findBestSplit(
+            const torch::Tensor& X,
+            const torch::Tensor& y,
+            const torch::Tensor& sample_weights
+        );
+
+        // Calculate weighted Gini impurity for multi-class
+        double calculateGiniImpurity(
+            const torch::Tensor& y,
+            const torch::Tensor& sample_weights
+        );
+
+
+
+        // Traverse tree to find leaf node
+        const TreeNode* traverseTree(const torch::Tensor& x, const TreeNode* node) const;
+
+        // Convert tree to graph representation
+        void treeToGraph(
+            const TreeNode* node,
+            std::vector<std::string>& lines,
+            int& node_id,
+            int parent_id = -1,
+            const std::string& edge_label = ""
+        ) const;
+    };
+
+    // Tree node structure
+    struct TreeNode {
+        bool is_leaf;
+
+        // For internal nodes
+        int split_feature;
+        int split_value;
+        std::unique_ptr<TreeNode> left;
+        std::unique_ptr<TreeNode> right;
+
+        // For leaf nodes
+        int predicted_class;
+        torch::Tensor class_probabilities;  // Probability for each class
+
+        TreeNode() : is_leaf(false), split_feature(-1), split_value(-1), predicted_class(-1) {}
+    };
+
+} // namespace bayesnet
+
+#endif // DECISION_TREE_H

src/experimental_clfs/ExpClf.cpp

@@ -0,0 +1,182 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include "ExpClf.h"
+#include "TensorUtils.hpp"
+
+namespace platform {
+    ExpClf::ExpClf() : semaphore_{ CountingSemaphore::getInstance() }, Boost(false)
+    {
+        validHyperparameters = {};
+    }
+    //
+    // Parents
+    //
+    void ExpClf::add_active_parents(const std::vector<int>& active_parents)
+    {
+        for (const auto& parent : active_parents)
+            aode_.add_active_parent(parent);
+    }
+    void ExpClf::add_active_parent(int parent)
+    {
+        aode_.add_active_parent(parent);
+    }
+    void ExpClf::remove_last_parent()
+    {
+        aode_.remove_last_parent();
+    }
+    //
+    // Predict
+    //
+    std::vector<int> ExpClf::predict_spode(std::vector<std::vector<int>>& test_data, int parent)
+    {
+        int test_size = test_data[0].size();
+        int sample_size = test_data.size();
+        auto predictions = std::vector<int>(test_size);
+
+        int chunk_size = std::min(150, int(test_size / semaphore_.getMaxCount()) + 1);
+        std::vector<std::thread> threads;
+        auto worker = [&](const std::vector<std::vector<int>>& samples, int begin, int chunk, int sample_size, std::vector<int>& predictions) {
+            std::string threadName = "(V)PWorker-" + std::to_string(begin) + "-" + std::to_string(chunk);
+#if defined(__linux__)
+            pthread_setname_np(pthread_self(), threadName.c_str());
+#else
+            pthread_setname_np(threadName.c_str());
+#endif
+            std::vector<int> instance(sample_size);
+            for (int sample = begin; sample < begin + chunk; ++sample) {
+                for (int feature = 0; feature < sample_size; ++feature) {
+                    instance[feature] = samples[feature][sample];
+                }
+                predictions[sample] = aode_.predict_spode(instance, parent);
+            }
+            semaphore_.release();
+            };
+        for (int begin = 0; begin < test_size; begin += chunk_size) {
+            int chunk = std::min(chunk_size, test_size - begin);
+            semaphore_.acquire();
+            threads.emplace_back(worker, test_data, begin, chunk, sample_size, std::ref(predictions));
+        }
+        for (auto& thread : threads) {
+            thread.join();
+        }
+        return predictions;
+    }
+    torch::Tensor ExpClf::predict(torch::Tensor& X)
+    {
+        auto X_ = TensorUtils::to_matrix(X);
+        torch::Tensor y = torch::tensor(predict(X_));
+        return y;
+    }
+    torch::Tensor ExpClf::predict_proba(torch::Tensor& X)
+    {
+        auto X_ = TensorUtils::to_matrix(X);
+        auto probabilities = predict_proba(X_);
+        auto n_samples = X.size(1);
+        int n_classes = probabilities[0].size();
+        auto y = torch::zeros({ n_samples, n_classes });
+        for (int i = 0; i < n_samples; i++) {
+            for (int j = 0; j < n_classes; j++) {
+                y[i][j] = probabilities[i][j];
+            }
+        }
+        return y;
+    }
+    float ExpClf::score(torch::Tensor& X, torch::Tensor& y)
+    {
+        auto X_ = TensorUtils::to_matrix(X);
+        auto y_ = TensorUtils::to_vector<int>(y);
+        return score(X_, y_);
+    }
+    std::vector<std::vector<double>> ExpClf::predict_proba(const std::vector<std::vector<int>>& test_data)
+    {
+        int test_size = test_data[0].size();
+        int sample_size = test_data.size();
+        auto probabilities = std::vector<std::vector<double>>(test_size, std::vector<double>(aode_.statesClass()));
+
+        int chunk_size = std::min(150, int(test_size / semaphore_.getMaxCount()) + 1);
+        std::vector<std::thread> threads;
+        auto worker = [&](const std::vector<std::vector<int>>& samples, int begin, int chunk, int sample_size, std::vector<std::vector<double>>& predictions) {
+            std::string threadName = "(V)PWorker-" + std::to_string(begin) + "-" + std::to_string(chunk);
+#if defined(__linux__)
+            pthread_setname_np(pthread_self(), threadName.c_str());
+#else
+            pthread_setname_np(threadName.c_str());
+#endif
+
+            std::vector<int> instance(sample_size);
+            for (int sample = begin; sample < begin + chunk; ++sample) {
+                for (int feature = 0; feature < sample_size; ++feature) {
+                    instance[feature] = samples[feature][sample];
+                }
+                predictions[sample] = aode_.predict_proba(instance);
+            }
+            semaphore_.release();
+            };
+        for (int begin = 0; begin < test_size; begin += chunk_size) {
+            int chunk = std::min(chunk_size, test_size - begin);
+            semaphore_.acquire();
+            threads.emplace_back(worker, test_data, begin, chunk, sample_size, std::ref(probabilities));
+        }
+        for (auto& thread : threads) {
+            thread.join();
+        }
+        return probabilities;
+    }
+    std::vector<int> ExpClf::predict(std::vector<std::vector<int>>& test_data)
+    {
+        if (!fitted) {
+            throw std::logic_error(CLASSIFIER_NOT_FITTED);
+        }
+        auto probabilities = predict_proba(test_data);
+        std::vector<int> predictions(probabilities.size(), 0);
+
+        for (size_t i = 0; i < probabilities.size(); i++) {
+            predictions[i] = std::distance(probabilities[i].begin(), std::max_element(probabilities[i].begin(), probabilities[i].end()));
+        }
+
+        return predictions;
+    }
+    float ExpClf::score(std::vector<std::vector<int>>& test_data, std::vector<int>& labels)
+    {
+        Timer timer;
+        timer.start();
+        std::vector<int> predictions = predict(test_data);
+        int correct = 0;
+
+        for (size_t i = 0; i < predictions.size(); i++) {
+            if (predictions[i] == labels[i]) {
+                correct++;
+            }
+        }
+        if (debug) {
+            std::cout << "* Time to predict: " << timer.getDurationString() << std::endl;
+        }
+        return static_cast<float>(correct) / predictions.size();
+    }
+
+    //
+    // statistics
+    //
+    int ExpClf::getNumberOfNodes() const
+    {
+        return aode_.getNumberOfNodes();
+    }
+    int ExpClf::getNumberOfEdges() const
+    {
+        return aode_.getNumberOfEdges();
+    }
+    int ExpClf::getNumberOfStates() const
+    {
+        return aode_.getNumberOfStates();
+    }
+    int ExpClf::getClassNumStates() const
+    {
+        return aode_.statesClass();
+    }
+
+
+}

src/experimental_clfs/ExpClf.h

@@ -0,0 +1,67 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#ifndef EXPCLF_H
+#define EXPCLF_H
+#include <vector>
+#include <string>
+#include <cmath>
+#include <algorithm>
+#include <limits>
+#include <bayesnet/ensembles/Boost.h>
+#include <bayesnet/network/Smoothing.h>
+#include "common/Timer.hpp"
+#include "CountingSemaphore.hpp"
+#include "Xaode.hpp"
+
+namespace platform {
+    class ExpClf : public bayesnet::Boost {
+    public:
+        ExpClf();
+        virtual ~ExpClf() = default;
+        std::vector<int> predict(std::vector<std::vector<int>>& X) override;
+        torch::Tensor predict(torch::Tensor& X) override;
+        torch::Tensor predict_proba(torch::Tensor& X) override;
+        std::vector<int> predict_spode(std::vector<std::vector<int>>& test_data, int parent);
+        std::vector<std::vector<double>> predict_proba(const std::vector<std::vector<int>>& X);
+        float score(std::vector<std::vector<int>>& X, std::vector<int>& y) override;
+        float score(torch::Tensor& X, torch::Tensor& y) override;
+        int getNumberOfNodes() const override;
+        int getNumberOfEdges() const override;
+        int getNumberOfStates() const override;
+        int getClassNumStates() const override;
+        std::vector<std::string> show() const override { return {}; }
+        std::vector<std::string> topological_order()  override { return {}; }
+        std::string dump_cpt() const override { return ""; }
+        void setDebug(bool debug) { this->debug = debug; }
+        bayesnet::status_t getStatus() const override { return status; }
+        std::vector<std::string> getNotes() const override { return notes; }
+        std::vector<std::string> graph(const std::string& title = "") const override { return {}; }
+        void add_active_parents(const std::vector<int>& active_parents);
+        void add_active_parent(int parent);
+        void remove_last_parent();
+        void setHyperparameters(const nlohmann::json& hyperparameters_) override {};
+    protected:
+        bool debug = false;
+        Xaode aode_;
+        torch::Tensor weights_;
+        const std::string CLASSIFIER_NOT_FITTED = "Classifier has not been fitted";
+        inline void normalize_weights(int num_instances)
+        {
+            double sum = weights_.sum().item<double>();
+            if (sum == 0) {
+                weights_ = torch::full({ num_instances }, 1.0);
+            } else {
+                for (int i = 0; i < weights_.size(0); ++i) {
+                    weights_[i] = weights_[i].item<double>() * num_instances / sum;
+                }
+            }
+        }
+    private:
+        CountingSemaphore& semaphore_;
+    };
+}
+#endif // EXPCLF_H

src/experimental_clfs/ExpEnsemble.cpp

@@ -0,0 +1,158 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include "ExpEnsemble.h"
+#include "TensorUtils.hpp"
+
+namespace platform {
+    ExpEnsemble::ExpEnsemble() : semaphore_{ CountingSemaphore::getInstance() }, Boost(false)
+    {
+        validHyperparameters = {};
+    }
+    //
+    // Parents
+    //
+    void ExpEnsemble::add_model(std::unique_ptr<XSpode> model)
+    {
+        models.push_back(std::move(model));
+        n_models++;
+    }
+    void ExpEnsemble::remove_last_model()
+    {
+        models.pop_back();
+        n_models--;
+    }
+    //
+    // Predict
+    //
+    torch::Tensor ExpEnsemble::predict(torch::Tensor& X)
+    {
+        auto X_ = TensorUtils::to_matrix(X);
+        torch::Tensor y = torch::tensor(predict(X_));
+        return y;
+    }
+    torch::Tensor ExpEnsemble::predict_proba(torch::Tensor& X)
+    {
+        auto X_ = TensorUtils::to_matrix(X);
+        auto probabilities = predict_proba(X_);
+        auto n_samples = X.size(1);
+        int n_classes = probabilities[0].size();
+        auto y = torch::zeros({ n_samples, n_classes });
+        for (int i = 0; i < n_samples; i++) {
+            for (int j = 0; j < n_classes; j++) {
+                y[i][j] = probabilities[i][j];
+            }
+        }
+        return y;
+    }
+    float ExpEnsemble::score(torch::Tensor& X, torch::Tensor& y)
+    {
+        auto X_ = TensorUtils::to_matrix(X);
+        auto y_ = TensorUtils::to_vector<int>(y);
+        return score(X_, y_);
+    }
+    std::vector<std::vector<double>> ExpEnsemble::predict_proba(const std::vector<std::vector<int>>& test_data)
+    {
+        int test_size = test_data[0].size();
+        int sample_size = test_data.size();
+        auto probabilities = std::vector<std::vector<double>>(test_size, std::vector<double>(getClassNumStates()));
+        int chunk_size = std::min(150, int(test_size / semaphore_.getMaxCount()) + 1);
+        std::vector<std::thread> threads;
+        auto worker = [&](const std::vector<std::vector<int>>& samples, int begin, int chunk, int sample_size, std::vector<std::vector<double>>& predictions) {
+            std::string threadName = "(V)PWorker-" + std::to_string(begin) + "-" + std::to_string(chunk);
+#if defined(__linux__)
+            pthread_setname_np(pthread_self(), threadName.c_str());
+#else
+            pthread_setname_np(threadName.c_str());
+#endif
+
+            std::vector<int> instance(sample_size);
+            for (int sample = begin; sample < begin + chunk; ++sample) {
+                for (int feature = 0; feature < sample_size; ++feature) {
+                    instance[feature] = samples[feature][sample];
+                }
+                // predictions[sample] = aode_.predict_proba(instance);
+            }
+            semaphore_.release();
+            };
+        for (int begin = 0; begin < test_size; begin += chunk_size) {
+            int chunk = std::min(chunk_size, test_size - begin);
+            semaphore_.acquire();
+            threads.emplace_back(worker, test_data, begin, chunk, sample_size, std::ref(probabilities));
+        }
+        for (auto& thread : threads) {
+            thread.join();
+        }
+        return probabilities;
+    }
+    std::vector<int> ExpEnsemble::predict(std::vector<std::vector<int>>& test_data)
+    {
+        if (!fitted) {
+            throw std::logic_error(CLASSIFIER_NOT_FITTED);
+        }
+        auto probabilities = predict_proba(test_data);
+        std::vector<int> predictions(probabilities.size(), 0);
+
+        for (size_t i = 0; i < probabilities.size(); i++) {
+            predictions[i] = std::distance(probabilities[i].begin(), std::max_element(probabilities[i].begin(), probabilities[i].end()));
+        }
+
+        return predictions;
+    }
+    float ExpEnsemble::score(std::vector<std::vector<int>>& test_data, std::vector<int>& labels)
+    {
+        Timer timer;
+        timer.start();
+        std::vector<int> predictions = predict(test_data);
+        int correct = 0;
+
+        for (size_t i = 0; i < predictions.size(); i++) {
+            if (predictions[i] == labels[i]) {
+                correct++;
+            }
+        }
+        if (debug) {
+            std::cout << "* Time to predict: " << timer.getDurationString() << std::endl;
+        }
+        return static_cast<float>(correct) / predictions.size();
+    }
+
+    //
+    // statistics
+    //
+    int ExpEnsemble::getNumberOfNodes() const
+    {
+        if (models_.empty()) {
+            return 0;
+        }
+        return n_models * (models_.at(0)->getNFeatures() + 1);
+    }
+    int ExpEnsemble::getNumberOfEdges() const
+    {
+        if (models_.empty()) {
+            return 0;
+        }
+        return n_models * (2 * models_.at(0)->getNFeatures() - 1);
+    }
+    int ExpEnsemble::getNumberOfStates() const
+    {
+        if (models_.empty()) {
+            return 0;
+        }
+        auto states = models_.at(0)->getStates();
+        int nFeatures = models_.at(0)->getNFeatures();
+        return std::accumulate(states.begin(), states.end(), 0) * nFeatures * n_models;
+    }
+    int ExpEnsemble::getClassNumStates() const
+    {
+        if (models_.empty()) {
+            return 0;
+        }
+        return models_.at(0)->statesClass();
+    }
+
+
+}

src/experimental_clfs/ExpEnsemble.h

@@ -0,0 +1,66 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#ifndef EXPENSEMBLE_H
+#define EXPENSEMBLE_H
+#include <vector>
+#include <string>
+#include <cmath>
+#include <algorithm>
+#include <limits>
+#include <bayesnet/ensembles/Boost.h>
+#include <bayesnet/network/Smoothing.h>
+#include "common/Timer.hpp"
+#include "CountingSemaphore.hpp"
+#include "XSpode.hpp"
+
+namespace platform {
+    class ExpEnsemble : public bayesnet::Boost {
+    public:
+        ExpEnsemble();
+        virtual ~ExpEnsemble() = default;
+        std::vector<int> predict(std::vector<std::vector<int>>& X) override;
+        torch::Tensor predict(torch::Tensor& X) override;
+        torch::Tensor predict_proba(torch::Tensor& X) override;
+        std::vector<int> predict_spode(std::vector<std::vector<int>>& test_data, int parent);
+        std::vector<std::vector<double>> predict_proba(const std::vector<std::vector<int>>& X);
+        float score(std::vector<std::vector<int>>& X, std::vector<int>& y) override;
+        float score(torch::Tensor& X, torch::Tensor& y) override;
+        int getNumberOfNodes() const override;
+        int getNumberOfEdges() const override;
+        int getNumberOfStates() const override;
+        int getClassNumStates() const override;
+        std::vector<std::string> show() const override { return {}; }
+        std::vector<std::string> topological_order()  override { return {}; }
+        std::string dump_cpt() const override { return ""; }
+        void setDebug(bool debug) { this->debug = debug; }
+        bayesnet::status_t getStatus() const override { return status; }
+        std::vector<std::string> getNotes() const override { return notes; }
+        std::vector<std::string> graph(const std::string& title = "") const override { return {}; }
+    protected:
+        void add_model(std::unique_ptr<XSpode> model);
+        void remove_last_model();
+        bool debug = false;
+        std::vector <std::unique_ptr<XSpode>> models_;
+        torch::Tensor weights_;
+        std::vector<double> significanceModels_;
+        const std::string CLASSIFIER_NOT_FITTED = "Classifier has not been fitted";
+        inline void normalize_weights(int num_instances)
+        {
+            double sum = weights_.sum().item<double>();
+            if (sum == 0) {
+                weights_ = torch::full({ num_instances }, 1.0);
+            } else {
+                for (int i = 0; i < weights_.size(0); ++i) {
+                    weights_[i] = weights_[i].item<double>() * num_instances / sum;
+                }
+            }
+        }
+    private:
+        CountingSemaphore& semaphore_;
+    };
+}
+#endif // EXPENSEMBLE_H

src/experimental_clfs/README.md

@@ -0,0 +1,142 @@
+# AdaBoost and DecisionTree Classifier Implementation
+
+This implementation provides both a Decision Tree classifier and a multi-class AdaBoost classifier based on the SAMME (Stagewise Additive Modeling using a Multi-class Exponential loss) algorithm described in the paper "Multi-class AdaBoost" by Zhu et al. Implemented in C++ using <https://claude.ai>
+
+## Components
+
+### 1. DecisionTree Classifier
+
+A classic decision tree implementation that:
+
+- Supports multi-class classification
+- Handles weighted samples (essential for boosting)
+- Uses Gini impurity as the splitting criterion
+- Works with discrete/categorical features
+- Provides both class predictions and probability estimates
+
+#### Key Features
+
+- **Max Depth Control**: Limit tree depth to create weak learners
+- **Minimum Samples**: Control minimum samples for splitting and leaf nodes
+- **Weighted Training**: Properly handles sample weights for boosting
+- **Visualization**: Generates DOT format graphs of the tree structure
+
+#### Hyperparameters
+
+- `max_depth`: Maximum depth of the tree (default: 3)
+- `min_samples_split`: Minimum samples required to split a node (default: 2)
+- `min_samples_leaf`: Minimum samples required in a leaf node (default: 1)
+
+### 2. AdaBoost Classifier
+
+A multi-class AdaBoost implementation using DecisionTree as base estimators:
+
+- **SAMME Algorithm**: Implements the multi-class extension of AdaBoost
+- **Automatic Stumps**: Uses decision stumps (max_depth=1) by default
+- **Early Stopping**: Stops if base classifier performs worse than random
+- **Ensemble Visualization**: Shows the weighted combination of base estimators
+
+#### Key Features
+
+- **Multi-class Support**: Natural extension to K classes
+- **Base Estimator Control**: Configure depth of base decision trees
+- **Training Monitoring**: Track training errors and estimator weights
+- **Probability Estimates**: Provides class probability predictions
+
+#### Hyperparameters
+
+- `n_estimators`: Number of base estimators to train (default: 50)
+- `base_max_depth`: Maximum depth for base decision trees (default: 1)
+
+## Algorithm Details
+
+The SAMME algorithm differs from binary AdaBoost in the calculation of the estimator weight (alpha):
+
+```
+α = log((1 - err) / err) + log(K - 1)
+```
+
+where `K` is the number of classes. This formula ensures that:
+
+- When K = 2, it reduces to standard AdaBoost
+- For K > 2, base classifiers only need to be better than random guessing (1/K) rather than 50%
+
+## Usage Example
+
+```cpp
+// Create AdaBoost with decision stumps
+AdaBoost ada(100, 1);  // 100 estimators, max_depth=1
+
+// Train
+ada.fit(X_train, y_train, features, className, states, Smoothing_t::NONE);
+
+// Predict
+auto predictions = ada.predict(X_test);
+auto probabilities = ada.predict_proba(X_test);
+
+// Evaluate
+float accuracy = ada.score(X_test, y_test);
+
+// Get ensemble information
+auto weights = ada.getEstimatorWeights();
+auto errors = ada.getTrainingErrors();
+```
+
+## Implementation Structure
+
+```
+AdaBoost (inherits from Ensemble)
+    └── Uses multiple DecisionTree instances as base estimators
+         └── DecisionTree (inherits from Classifier)
+              └── Implements weighted Gini impurity splitting
+```
+
+## Visualization
+
+Both classifiers support graph visualization:
+
+- **DecisionTree**: Shows the tree structure with split conditions
+- **AdaBoost**: Shows the ensemble of weighted base estimators
+
+Generate visualizations using:
+
+```cpp
+auto graph = classifier.graph("Title");
+```
+
+## Data Format
+
+Both classifiers expect discrete/categorical data:
+
+- **Features**: Integer values representing categories (stored in `torch::Tensor` or `std::vector<std::vector<int>>`)
+- **Labels**: Integer values representing class indices (0, 1, ..., K-1)
+- **States**: Map defining possible values for each feature and the class variable
+- **Sample Weights**: Optional weights for each training sample (important for boosting)
+
+Example data setup:
+
+```cpp
+// Features matrix (n_features x n_samples)
+torch::Tensor X = torch::tensor({{0, 1, 2}, {1, 0, 1}});  // 2 features, 3 samples
+
+// Labels vector
+torch::Tensor y = torch::tensor({0, 1, 0});  // 3 samples
+
+// States definition
+std::map<std::string, std::vector<int>> states;
+states["feature1"] = {0, 1, 2};  // Feature 1 can take values 0, 1, or 2
+states["feature2"] = {0, 1};     // Feature 2 can take values 0 or 1
+states["class"] = {0, 1};        // Binary classification
+```
+
+## Notes
+
+- The implementation handles discrete/categorical features as indicated by the int-based data structures
+- Sample weights are properly propagated through the tree building process
+- The DecisionTree implementation uses equality testing for splits (suitable for categorical data)
+- Both classifiers support the standard fit/predict interface from the base framework
+
+## References
+
+- Zhu, J., Zou, H., Rosset, S., & Hastie, T. (2009). Multi-class AdaBoost. Statistics and its interface, 2(3), 349-360.
+- Breiman, L., Friedman, J., Olshen, R., & Stone, C. (1984). Classification and Regression Trees. Wadsworth, Belmont, CA.

src/experimental_clfs/TensorUtils.hpp

@@ -0,0 +1,97 @@
+#ifndef TENSORUTILS_HPP
+#define TENSORUTILS_HPP
+#include <torch/torch.h>
+#include <vector>
+namespace platform {
+    class TensorUtils {
+    public:
+        static std::vector<std::vector<int>> to_matrix(const torch::Tensor& X)
+        {
+            // Ensure tensor is contiguous in memory
+            auto X_contig = X.contiguous();
+
+            // Access tensor data pointer directly
+            auto data_ptr = X_contig.data_ptr<int>();
+
+            // IF you are using int64_t as the data type, use the following line
+            //auto data_ptr = X_contig.data_ptr<int64_t>();
+            //std::vector<std::vector<int64_t>> data(X.size(0), std::vector<int64_t>(X.size(1)));
+
+            // Prepare output container
+            std::vector<std::vector<int>> data(X.size(0), std::vector<int>(X.size(1)));
+
+            // Fill the 2D vector in a single loop using pointer arithmetic
+            int rows = X.size(0);
+            int cols = X.size(1);
+            for (int i = 0; i < rows; ++i) {
+                std::copy(data_ptr + i * cols, data_ptr + (i + 1) * cols, data[i].begin());
+            }
+            return data;
+        }
+        template <typename T>
+        static std::vector<T> to_vector(const torch::Tensor& y)
+        {
+            // Ensure the tensor is contiguous in memory
+            auto y_contig = y.contiguous();
+
+            // Access data pointer
+            auto data_ptr = y_contig.data_ptr<T>();
+
+            // Prepare output container
+            std::vector<T> data(y.size(0));
+
+            // Copy data efficiently
+            std::copy(data_ptr, data_ptr + y.size(0), data.begin());
+
+            return data;
+        }
+        static torch::Tensor to_matrix(const std::vector<std::vector<int>>& data)
+        {
+            if (data.empty()) return torch::empty({ 0, 0 }, torch::kInt64);
+            size_t rows = data.size();
+            size_t cols = data[0].size();
+            torch::Tensor tensor = torch::empty({ static_cast<long>(rows), static_cast<long>(cols) }, torch::kInt64);
+            for (size_t i = 0; i < rows; ++i) {
+                for (size_t j = 0; j < cols; ++j) {
+                    tensor.index_put_({ static_cast<long>(i), static_cast<long>(j) }, data[i][j]);
+                }
+            }
+            return tensor;
+        }
+    };
+    static void dumpVector(const std::vector<std::vector<int>>& vec, const std::string& name)
+    {
+        std::cout << name << ": " << std::endl;
+        for (const auto& row : vec) {
+            std::cout << "[";
+            for (const auto& val : row) {
+                std::cout << val << " ";
+            }
+            std::cout << "]" << std::endl;
+        }
+        std::cout << std::endl;
+    }
+    static void dumpTensor(const torch::Tensor& tensor, const std::string& name)
+    {
+        std::cout << name << ": " << std::endl;
+        for (auto i = 0; i < tensor.size(0); i++) {
+            std::cout << "[";
+            for (auto j = 0; j < tensor.size(1); j++) {
+                std::cout << tensor[i][j].item<int>() << " ";
+            }
+            std::cout << "]" << std::endl;
+        }
+        std::cout << std::endl;
+    }
+    static void dumpTensorV(const torch::Tensor& tensor, const std::string& name)
+    {
+        std::cout << name << ": " << std::endl;
+        std::cout << "[";
+        for (int i = 0; i < tensor.size(0); i++) {
+            std::cout << tensor[i].item<int>() << " ";
+        }
+        std::cout << "]" << std::endl;
+    }
+}
+
+#endif // TENSORUTILS_HPP

src/experimental_clfs/XA1DE.cpp

@@ -0,0 +1,20 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include "XA1DE.h"
+#include "TensorUtils.hpp"
+
+namespace platform {
+    void XA1DE::trainModel(const torch::Tensor& weights, const bayesnet::Smoothing_t smoothing)
+    {
+        auto X = TensorUtils::to_matrix(dataset.slice(0, 0, dataset.size(0) - 1));
+        auto y = TensorUtils::to_vector<int>(dataset.index({ -1, "..." }));
+        int num_instances = X[0].size();
+        weights_ = torch::full({ num_instances }, 1.0);
+        //normalize_weights(num_instances);
+        aode_.fit(X, y, features, className, states, weights_, true, smoothing);
+    }
+}

src/experimental_clfs/XA1DE.h

@@ -0,0 +1,26 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#ifndef XA1DE_H
+#define XA1DE_H
+#include "Xaode.hpp"
+#include "ExpClf.h"
+#include <bayesnet/network/Smoothing.h>
+
+namespace platform {
+    class XA1DE : public ExpClf {
+    public:
+        XA1DE() = default;
+        virtual ~XA1DE() override = default;
+        std::string getVersion() override { return version; };
+    protected:
+        void buildModel(const torch::Tensor& weights) override {};
+        void trainModel(const torch::Tensor& weights, const bayesnet::Smoothing_t smoothing) override;
+    private:
+        std::string version = "1.0.0";
+    };
+}
+#endif // XA1DE_H

src/experimental_clfs/XBAODE.cpp

@@ -0,0 +1,183 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+#include <random> 
+#include <set>
+#include <functional>
+#include <limits.h>
+#include <tuple>
+#include "XBAODE.h"
+#include "XSpode.hpp"
+#include "TensorUtils.hpp"
+#include <loguru.hpp>
+
+namespace platform {
+    XBAODE::XBAODE()
+    {
+        validHyperparameters = { "alpha_block", "order", "convergence", "convergence_best", "bisection", "threshold", "maxTolerance",
+            "predict_voting", "select_features" };
+    }
+    void XBAODE::add_model(std::unique_ptr<XSpode> model)
+    {
+        models.push_back(std::move(model));
+        n_models++;
+    }
+    void XBAODE::remove_last_model()
+    {
+        models.pop_back();
+        n_models--;
+    }
+    void XBAODE::trainModel(const torch::Tensor& weights, const bayesnet::Smoothing_t smoothing)
+    {
+        fitted = true;
+        X_train_ = TensorUtils::to_matrix(X_train);
+        y_train_ = TensorUtils::to_vector<int>(y_train);
+        X_test_ = TensorUtils::to_matrix(X_test);
+        y_test_ = TensorUtils::to_vector<int>(y_test);
+        maxTolerance = 3;
+        //
+        // Logging setup
+        //
+        // loguru::set_thread_name("XBAODE");
+        // loguru::g_stderr_verbosity = loguru::Verbosity_OFF;
+        // loguru::add_file("XBAODE.log", loguru::Truncate, loguru::Verbosity_MAX);
+
+        // Algorithm based on the adaboost algorithm for classification
+        // as explained in Ensemble methods (Zhi-Hua Zhou, 2012)
+        double alpha_t = 0;
+        weights_ = torch::full({ m }, 1.0 / static_cast<double>(m), torch::kFloat64); // m initialized in Classifier.cc
+        significanceModels.resize(n, 0.0); // n initialized in Classifier.cc
+        bool finished = false;
+        std::vector<int> featuresUsed;
+        n_models = 0;
+        std::unique_ptr<XSpode> model;
+        if (selectFeatures) {
+            featuresUsed = featureSelection(weights_);
+            for (const auto& parent : featuresUsed) {
+                model = std::unique_ptr<XSpode>(new XSpode(parent));
+                model->fit(X_train_, y_train_, weights_, smoothing);
+                std::cout << model->getNFeatures() << std::endl;
+                add_model(std::move(model));
+            }
+            notes.push_back("Used features in initialization: " + std::to_string(featuresUsed.size()) + " of " + std::to_string(features.size()) + " with " + select_features_algorithm);
+            auto ypred = ExpEnsemble::predict(X_train);
+            std::tie(weights_, alpha_t, finished) = update_weights(y_train, ypred, weights_);
+            // Update significance of the models
+            for (const auto& parent : featuresUsed) {
+                significanceModels_[parent] = alpha_t;
+            }
+            n_models = featuresUsed.size();
+            // VLOG_SCOPE_F(1, "SelectFeatures. alpha_t: %f n_models: %d", alpha_t, n_models);
+            if (finished) {
+                return;
+            }
+        }
+        int numItemsPack = 0; // The counter of the models inserted in the current pack
+        // Variables to control the accuracy finish condition
+        double priorAccuracy = 0.0;
+        double improvement = 1.0;
+        double convergence_threshold = 1e-4;
+        int tolerance = 0; // number of times the accuracy is lower than the convergence_threshold
+        // Step 0: Set the finish condition
+        // epsilon sub t > 0.5 => inverse the weights policy
+        // validation error is not decreasing
+        // run out of features
+        bool ascending = order_algorithm == bayesnet::Orders.ASC;
+        std::mt19937 g{ 173 };
+        while (!finished) {
+            // Step 1: Build ranking with mutual information
+            auto featureSelection = metrics.SelectKBestWeighted(weights_, ascending, n); // Get all the features sorted
+            if (order_algorithm == bayesnet::Orders.RAND) {
+                std::shuffle(featureSelection.begin(), featureSelection.end(), g);
+            }
+            // Remove used features
+            featureSelection.erase(remove_if(featureSelection.begin(), featureSelection.end(), [&](auto x)
+                { return std::find(featuresUsed.begin(), featuresUsed.end(), x) != featuresUsed.end();}),
+                featureSelection.end()
+            );
+            int k = bisection ? pow(2, tolerance) : 1;
+            int counter = 0; // The model counter of the current pack
+            // VLOG_SCOPE_F(1, "counter=%d k=%d featureSelection.size: %zu", counter, k, featureSelection.size());
+            while (counter++ < k && featureSelection.size() > 0) {
+                auto feature = featureSelection[0];
+                featureSelection.erase(featureSelection.begin());
+                model = std::unique_ptr<XSpode>(new XSpode(feature));
+                model->fit(X_train_, y_train_, weights_, smoothing);
+                std::vector<int> ypred;
+                if (alpha_block) {
+                    //
+                    // Compute the prediction with the current ensemble + model
+                    //
+                    // Add the model to the ensemble
+                    significanceModels[feature] = 1.0;
+                    add_model(std::move(model));
+                    // Compute the prediction
+                    ypred = ExpEnsemble::predict(X_train_);
+                    // Remove the model from the ensemble
+                    significanceModels[feature] = 0.0;
+                    model = std::move(models_.back());
+                    remove_last_model();
+                } else {
+                    ypred = model->predict(X_train_);
+                }
+                // Step 3.1: Compute the classifier amout of say
+                auto ypred_t = torch::tensor(ypred);
+                std::tie(weights_, alpha_t, finished) = update_weights(y_train, ypred_t, weights_);
+                // Step 3.4: Store classifier and its accuracy to weigh its future vote
+                numItemsPack++;
+                featuresUsed.push_back(feature);
+                add_model(std::move(model));
+                significanceModels[feature] = alpha_t;
+                // VLOG_SCOPE_F(2, "finished: %d numItemsPack: %d n_models: %d featuresUsed: %zu", finished, numItemsPack, n_models, featuresUsed.size());
+            } // End of the pack
+            if (convergence && !finished) {
+                auto y_val_predict = ExpEnsemble::predict(X_test);
+                double accuracy = (y_val_predict == y_test).sum().item<double>() / (double)y_test.size(0);
+                if (priorAccuracy == 0) {
+                    priorAccuracy = accuracy;
+                } else {
+                    improvement = accuracy - priorAccuracy;
+                }
+                if (improvement < convergence_threshold) {
+                    // VLOG_SCOPE_F(3, "  (improvement<threshold) tolerance: %d numItemsPack: %d improvement: %f prior: %f current: %f", tolerance, numItemsPack, improvement, priorAccuracy, accuracy);
+                    tolerance++;
+                } else {
+                    // VLOG_SCOPE_F(3, "* (improvement>=threshold) Reset. tolerance: %d numItemsPack: %d improvement: %f prior: %f current: %f", tolerance, numItemsPack, improvement, priorAccuracy, accuracy);
+                    tolerance = 0; // Reset the counter if the model performs better
+                    numItemsPack = 0;
+                }
+                if (convergence_best) {
+                    // Keep the best accuracy until now as the prior accuracy
+                    priorAccuracy = std::max(accuracy, priorAccuracy);
+                } else {
+                    // Keep the last accuray obtained as the prior accuracy
+                    priorAccuracy = accuracy;
+                }
+            }
+            // VLOG_SCOPE_F(1, "tolerance: %d featuresUsed.size: %zu features.size: %zu", tolerance, featuresUsed.size(), features.size());
+            finished = finished || tolerance > maxTolerance || featuresUsed.size() == features.size();
+        }
+        if (tolerance > maxTolerance) {
+            if (numItemsPack < n_models) {
+                notes.push_back("Convergence threshold reached & " + std::to_string(numItemsPack) + " models eliminated");
+                // VLOG_SCOPE_F(4, "Convergence threshold reached & %d models eliminated of %d", numItemsPack, n_models);
+                for (int i = featuresUsed.size() - 1; i >= featuresUsed.size() - numItemsPack; --i) {
+                    remove_last_model();
+                    significanceModels[featuresUsed[i]] = 0.0;
+                }
+                // VLOG_SCOPE_F(4, "*Convergence threshold %d models left & %d features used.", n_models, featuresUsed.size());
+            } else {
+                notes.push_back("Convergence threshold reached & 0 models eliminated");
+                // VLOG_SCOPE_F(4, "Convergence threshold reached & 0 models eliminated n_models=%d numItemsPack=%d", n_models, numItemsPack);
+            }
+        }
+        if (featuresUsed.size() != features.size()) {
+            notes.push_back("Used features in train: " + std::to_string(featuresUsed.size()) + " of " + std::to_string(features.size()));
+            status = bayesnet::WARNING;
+        }
+        notes.push_back("Number of models: " + std::to_string(n_models));
+        return;
+    }
+}

src/experimental_clfs/XBAODE.h

@@ -0,0 +1,35 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#ifndef XBAODE_H
+#define XBAODE_H
+#include <iostream>
+#include <vector>
+#include <cmath>
+#include <algorithm>
+#include <limits>
+#include "common/Timer.hpp"
+#include "ExpEnsemble.h"
+
+namespace platform {
+    class XBAODE : public Boost {
+
+        // Hay que hacer un vector de modelos entrenados y hacer un predict ensemble con todos ellos
+        // Probar XA1DE con smooth original y laplace y comprobar diferencias si se pasan pesos a 1 o a 1/m
+    public:
+        XBAODE();
+        std::string getVersion() override { return version; };
+    protected:
+        void trainModel(const torch::Tensor& weights, const bayesnet::Smoothing_t smoothing) override;
+    private:
+        void add_model(std::unique_ptr<XSpode> model);
+        void remove_last_model();
+        std::vector<std::vector<int>> X_train_, X_test_;
+        std::vector<int> y_train_, y_test_;
+        std::string version = "0.9.7";
+    };
+}
+#endif // XBAODE_H

src/experimental_clfs/XSpode.hpp

@@ -0,0 +1,436 @@
+#ifndef XSPODE_H
+#define XSPODE_H
+
+#include <vector>
+#include <map>
+#include <stdexcept>
+#include <algorithm>
+#include <numeric>
+#include <string>
+#include <cmath>
+#include <limits>
+#include <sstream>
+#include <iostream>
+#include <torch/torch.h>
+#include <bayesnet/network/Smoothing.h>
+#include <bayesnet/classifiers/Classifier.h>
+#include "CountingSemaphore.hpp"
+
+
+namespace platform {
+
+    class XSpode : public bayesnet::Classifier {
+    public:
+        // --------------------------------------
+        // Constructor
+        //
+        // Supply which feature index is the single super-parent (“spIndex”).
+        // --------------------------------------
+        explicit XSpode(int spIndex)
+            : superParent_{ spIndex },
+            nFeatures_{ 0 },
+            statesClass_{ 0 },
+            fitted_{ false },
+            alpha_{ 1.0 },
+            initializer_{ 1.0 },
+            semaphore_{ CountingSemaphore::getInstance() } : bayesnet::Classifier(bayesnet::Network())
+        {
+        }
+
+        // --------------------------------------
+        // fit
+        // --------------------------------------
+        //
+        // Trains the SPODE given data:
+        //   X: X[f][n] is the f-th feature value for instance n
+        //   y: y[n] is the class value for instance n
+        //   states: a map or array that tells how many distinct states each feature and the class can take
+        //
+        // For example, states_.back() is the number of class states,
+        // and states_[f] is the number of distinct values for feature f.
+        //
+        // We only store conditional probabilities for:
+        //   p(x_sp| c)   (the super-parent feature)
+        //   p(x_child| c, x_sp)  for all child ≠ sp
+        //
+        // The “weights” can be a vector of per-instance weights; if not used, pass them as 1.0.
+        // --------------------------------------
+        void fit(const std::vector<std::vector<int>>& X,
+            const std::vector<int>& y,
+            const torch::Tensor& weights, const bayesnet::Smoothing_t smoothing)
+        {
+            int numInstances = static_cast<int>(y.size());
+            nFeatures_ = static_cast<int>(X.size());
+
+            // Derive the number of states for each feature and for the class.
+            // (This is just one approach; adapt to match your environment.)
+            // Here, we assume the user also gave us the total #states per feature in e.g. statesMap.
+            // We'll simply reconstruct the integer states_ array. The last entry is statesClass_.
+            states_.resize(nFeatures_);
+            for (int f = 0; f < nFeatures_; f++) {
+                // Suppose you look up in “statesMap” by the feature name, or read directly from X.
+                // We'll assume states_[f] = max value in X[f] + 1.
+                auto maxIt = std::max_element(X[f].begin(), X[f].end());
+                states_[f] = (*maxIt) + 1;
+            }
+            // For the class: states_.back() = max(y)+1
+            statesClass_ = (*std::max_element(y.begin(), y.end())) + 1;
+
+            // Initialize counts
+            classCounts_.resize(statesClass_, 0.0);
+            // p(x_sp = spVal | c)
+            // We'll store these counts in spFeatureCounts_[spVal * statesClass_ + c].
+            spFeatureCounts_.resize(states_[superParent_] * statesClass_, 0.0);
+
+            // For each child ≠ sp, we store p(childVal| c, spVal) in a separate block of childCounts_.
+            // childCounts_ will be sized as sum_{child≠sp} (states_[child] * statesClass_ * states_[sp]).
+            // We also need an offset for each child to index into childCounts_.
+            childOffsets_.resize(nFeatures_, -1);
+            int totalSize = 0;
+            for (int f = 0; f < nFeatures_; f++) {
+                if (f == superParent_) continue; // skip sp
+                childOffsets_[f] = totalSize;
+                // block size for this child's counts: states_[f] * statesClass_ * states_[superParent_]
+                totalSize += (states_[f] * statesClass_ * states_[superParent_]);
+            }
+            childCounts_.resize(totalSize, 0.0);
+
+            // Accumulate raw counts
+            for (int n = 0; n < numInstances; n++) {
+                std::vector<int> instance(nFeatures_ + 1);
+                for (int f = 0; f < nFeatures_; f++) {
+                    instance[f] = X[f][n];
+                }
+                instance[nFeatures_] = y[n];
+                addSample(instance, weights[n].item<double>());
+            }
+
+            switch (smoothing) {
+                case bayesnet::Smoothing_t::ORIGINAL:
+                    alpha_ = 1.0 / numInstances;
+                    break;
+                case bayesnet::Smoothing_t::LAPLACE:
+                    alpha_ = 1.0;
+                    break;
+                default:
+                    alpha_ = 0.0; // No smoothing 
+            }
+            initializer_ = initializer_ = std::numeric_limits<double>::max() / (nFeatures_ * nFeatures_);
+            // Convert raw counts to probabilities
+            computeProbabilities();
+            fitted_ = true;
+        }
+
+        // --------------------------------------
+        // addSample (only valid in COUNTS mode)
+        // --------------------------------------
+        //
+        // instance has size nFeatures_ + 1, with the class at the end.
+        // We add 1 to the appropriate counters for each (c, superParentVal, childVal).
+        //
+        void addSample(const std::vector<int>& instance, double weight)
+        {
+            if (weight <= 0.0) return;
+
+            int c = instance.back();
+            // (A) increment classCounts
+            classCounts_[c] += weight;
+
+            // (B) increment super-parent counts => p(x_sp | c)
+            int spVal = instance[superParent_];
+            spFeatureCounts_[spVal * statesClass_ + c] += weight;
+
+            // (C) increment child counts => p(childVal | c, x_sp)
+            for (int f = 0; f < nFeatures_; f++) {
+                if (f == superParent_) continue;
+                int childVal = instance[f];
+                int offset = childOffsets_[f];
+                // Compute index in childCounts_.
+                // Layout: [ offset + (spVal * states_[f] + childVal) * statesClass_ + c ]
+                int blockSize = states_[f] * statesClass_;
+                int idx = offset + spVal * blockSize + childVal * statesClass_ + c;
+                childCounts_[idx] += weight;
+            }
+        }
+
+        // --------------------------------------
+        // computeProbabilities
+        // --------------------------------------
+        //
+        // Once all samples are added in COUNTS mode, call this to:
+        //    p(c)
+        //    p(x_sp = spVal | c)
+        //    p(x_child = v | c, x_sp = s_sp)
+        //
+        // We store them in the corresponding *Probs_ arrays for inference.
+        // --------------------------------------
+        void computeProbabilities()
+        {
+            double totalCount = std::accumulate(classCounts_.begin(), classCounts_.end(), 0.0);
+
+            // p(c) => classPriors_
+            classPriors_.resize(statesClass_, 0.0);
+            if (totalCount <= 0.0) {
+                // fallback => uniform
+                double unif = 1.0 / static_cast<double>(statesClass_);
+                for (int c = 0; c < statesClass_; c++) {
+                    classPriors_[c] = unif;
+                }
+            } else {
+                for (int c = 0; c < statesClass_; c++) {
+                    classPriors_[c] = (classCounts_[c] + alpha_)
+                        / (totalCount + alpha_ * statesClass_);
+                }
+            }
+
+            // p(x_sp | c)
+            spFeatureProbs_.resize(spFeatureCounts_.size());
+            // denominator for spVal * statesClass_ + c is just classCounts_[c] + alpha_ * (#states of sp)
+            int spCard = states_[superParent_];
+            for (int spVal = 0; spVal < spCard; spVal++) {
+                for (int c = 0; c < statesClass_; c++) {
+                    double denom = classCounts_[c] + alpha_ * spCard;
+                    double num = spFeatureCounts_[spVal * statesClass_ + c] + alpha_;
+                    spFeatureProbs_[spVal * statesClass_ + c] = (denom <= 0.0 ? 0.0 : num / denom);
+                }
+            }
+
+            // p(x_child | c, x_sp)
+            childProbs_.resize(childCounts_.size());
+            for (int f = 0; f < nFeatures_; f++) {
+                if (f == superParent_) continue;
+                int offset = childOffsets_[f];
+                int childCard = states_[f];
+
+                // For each spVal, c, childVal in childCounts_:
+                for (int spVal = 0; spVal < spCard; spVal++) {
+                    for (int childVal = 0; childVal < childCard; childVal++) {
+                        for (int c = 0; c < statesClass_; c++) {
+                            int idx = offset + spVal * (childCard * statesClass_)
+                                + childVal * statesClass_
+                                + c;
+
+                            double num = childCounts_[idx] + alpha_;
+                            // denominator = spFeatureCounts_[spVal * statesClass_ + c] + alpha_ * (#states of child)
+                            double denom = spFeatureCounts_[spVal * statesClass_ + c]
+                                + alpha_ * childCard;
+                            childProbs_[idx] = (denom <= 0.0 ? 0.0 : num / denom);
+                        }
+                    }
+                }
+            }
+
+        }
+
+        // --------------------------------------
+        // predict_proba
+        // --------------------------------------
+        //
+        // For a single instance x of dimension nFeatures_:
+        //  P(c | x) ∝ p(c) × p(x_sp | c) × ∏(child ≠ sp) p(x_child | c, x_sp).
+        //
+        // Then we normalize the result.
+        // --------------------------------------
+        std::vector<double> predict_proba(const std::vector<int>& instance) const
+        {
+            std::vector<double> probs(statesClass_, 0.0);
+
+            // Multiply p(c) × p(x_sp | c)
+            int spVal = instance[superParent_];
+            for (int c = 0; c < statesClass_; c++) {
+                double pc = classPriors_[c];
+                double pSpC = spFeatureProbs_[spVal * statesClass_ + c];
+                probs[c] = pc * pSpC * initializer_;
+            }
+
+            // Multiply by each child’s probability p(x_child | c, x_sp)
+            for (int feature = 0; feature < nFeatures_; feature++) {
+                if (feature == superParent_) continue;  // skip sp
+                int sf = instance[feature];
+                int offset = childOffsets_[feature];
+                int childCard = states_[feature]; // not used directly, but for clarity
+                // Index into childProbs_ = offset + spVal*(childCard*statesClass_) + childVal*statesClass_ + c
+                int base = offset + spVal * (childCard * statesClass_) + sf * statesClass_;
+                for (int c = 0; c < statesClass_; c++) {
+                    probs[c] *= childProbs_[base + c];
+                }
+            }
+
+            // Normalize
+            normalize(probs);
+            return probs;
+        }
+        std::vector<std::vector<double>> predict_proba(const std::vector<std::vector<int>>& test_data)
+        {
+            int test_size = test_data[0].size();
+            int sample_size = test_data.size();
+            auto probabilities = std::vector<std::vector<double>>(test_size, std::vector<double>(statesClass_));
+
+            int chunk_size = std::min(150, int(test_size / semaphore_.getMaxCount()) + 1);
+            std::vector<std::thread> threads;
+            auto worker = [&](const std::vector<std::vector<int>>& samples, int begin, int chunk, int sample_size, std::vector<std::vector<double>>& predictions) {
+                std::string threadName = "(V)PWorker-" + std::to_string(begin) + "-" + std::to_string(chunk);
+#if defined(__linux__)
+                pthread_setname_np(pthread_self(), threadName.c_str());
+#else
+                pthread_setname_np(threadName.c_str());
+#endif
+
+                std::vector<int> instance(sample_size);
+                for (int sample = begin; sample < begin + chunk; ++sample) {
+                    for (int feature = 0; feature < sample_size; ++feature) {
+                        instance[feature] = samples[feature][sample];
+                    }
+                    predictions[sample] = predict_proba(instance);
+                }
+                semaphore_.release();
+                };
+            for (int begin = 0; begin < test_size; begin += chunk_size) {
+                int chunk = std::min(chunk_size, test_size - begin);
+                semaphore_.acquire();
+                threads.emplace_back(worker, test_data, begin, chunk, sample_size, std::ref(probabilities));
+            }
+            for (auto& thread : threads) {
+                thread.join();
+            }
+            return probabilities;
+        }
+
+        // --------------------------------------
+        // predict
+        // --------------------------------------
+        //
+        // Return the class argmax( P(c|x) ).
+        // --------------------------------------
+        int predict(const std::vector<int>& instance) const
+        {
+            auto p = predict_proba(instance);
+            return static_cast<int>(std::distance(p.begin(),
+                std::max_element(p.begin(), p.end())));
+        }
+        std::vector<int> predict(std::vector<std::vector<int>>& test_data)
+        {
+            if (!fitted_) {
+                throw std::logic_error(CLASSIFIER_NOT_FITTED);
+            }
+            auto probabilities = predict_proba(test_data);
+            std::vector<int> predictions(probabilities.size(), 0);
+
+            for (size_t i = 0; i < probabilities.size(); i++) {
+                predictions[i] = std::distance(probabilities[i].begin(), std::max_element(probabilities[i].begin(), probabilities[i].end()));
+            }
+
+            return predictions;
+        }
+
+        // --------------------------------------
+        // Utility: normalize
+        // --------------------------------------
+        void normalize(std::vector<double>& v) const
+        {
+            double sum = 0.0;
+            for (auto val : v) { sum += val; }
+            if (sum <= 0.0) {
+                return;
+            }
+            for (auto& val : v) {
+                val /= sum;
+            }
+        }
+
+        // --------------------------------------
+        // debug printing, if desired
+        // --------------------------------------
+        std::string to_string() const
+        {
+            std::ostringstream oss;
+            oss << "---- SPODE Model ----\n"
+                << "nFeatures_  = " << nFeatures_ << "\n"
+                << "superParent_ = " << superParent_ << "\n"
+                << "statesClass_ = " << statesClass_ << "\n"
+                << "\n";
+
+            oss << "States: [";
+            for (int s : states_) oss << s << " ";
+            oss << "]\n";
+
+            oss << "classCounts_: [";
+            for (double c : classCounts_) oss << c << " ";
+            oss << "]\n";
+
+            oss << "classPriors_: [";
+            for (double c : classPriors_) oss << c << " ";
+            oss << "]\n";
+
+            oss << "spFeatureCounts_: size = " << spFeatureCounts_.size() << "\n[";
+            for (double c : spFeatureCounts_) oss << c << " ";
+            oss << "]\n";
+
+            oss << "spFeatureProbs_: size = " << spFeatureProbs_.size() << "\n[";
+            for (double c : spFeatureProbs_) oss << c << " ";
+            oss << "]\n";
+
+            oss << "childCounts_: size = " << childCounts_.size() << "\n[";
+            for (double cc : childCounts_) oss << cc << " ";
+            oss << "]\n";
+
+            oss << "childProbs_: size = " << childProbs_.size() << "\n[";
+            for (double cp : childProbs_) oss << cp << " ";
+            oss << "]\n";
+
+            oss << "childOffsets_: [";
+            for (int co : childOffsets_) oss << co << " ";
+            oss << "]\n";
+
+            oss << "---------------------\n";
+            return oss.str();
+        }
+        int statesClass() const { return statesClass_; }
+        int getNFeatures() const { return nFeatures_; }
+        int getNumberOfStates() const
+        {
+            return std::accumulate(states_.begin(), states_.end(), 0) * nFeatures_;
+        }
+        int getNumberOfEdges() const
+        {
+            return nFeatures_ * (2 * nFeatures_ - 1);
+        }
+        std::vector<int>& getStates() { return states_; }
+
+    private:
+        // --------------------------------------
+        // MEMBERS
+        // --------------------------------------
+
+        int superParent_;                  // which feature is the single super-parent
+        int nFeatures_;
+        int statesClass_;
+        bool fitted_ = false;
+        std::vector<int> states_;          // [states_feat0, ..., states_feat(N-1)] (class not included in this array)
+
+        const std::string CLASSIFIER_NOT_FITTED = "Classifier has not been fitted";
+
+        // Class counts
+        std::vector<double> classCounts_;  // [c], accumulative
+        std::vector<double> classPriors_;  // [c], after normalization
+
+        // For p(x_sp = spVal | c)
+        std::vector<double> spFeatureCounts_; // [spVal * statesClass_ + c]
+        std::vector<double> spFeatureProbs_;  // same shape, after normalization
+
+        // For p(x_child = childVal | x_sp = spVal, c)
+        // childCounts_ is big enough to hold all child features except sp:
+        //   For each child f, we store childOffsets_[f] as the start index, then
+        //   childVal, spVal, c => the data.
+        std::vector<double> childCounts_;
+        std::vector<double> childProbs_;
+        std::vector<int>    childOffsets_;
+
+        double alpha_ = 1.0;
+        double initializer_; // for numerical stability
+        CountingSemaphore& semaphore_;
+    };
+
+} // namespace platform
+
+#endif // XSPODE_H

src/experimental_clfs/Xaode.hpp

@@ -0,0 +1,478 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2025 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+// Based on the Geoff. I. Webb A1DE java algorithm
+// https://weka.sourceforge.io/packageMetaData/AnDE/Latest.html
+
+#ifndef XAODE_H
+#define XAODE_H
+#include <vector>
+#include <map>
+#include <stdexcept>
+#include <algorithm>
+#include <numeric>
+#include <string>
+#include <cmath>
+#include <limits>
+#include <sstream>
+#include <torch/torch.h>
+#include <bayesnet/network/Smoothing.h>
+
+
+namespace platform {
+    class Xaode {
+    public:
+        // -------------------------------------------------------
+        // The Xaode can be EMPTY (just created), in COUNTS mode (accumulating raw counts)
+        // or PROBS mode (storing conditional probabilities).
+        enum class MatrixState {
+            EMPTY,
+            COUNTS,
+            PROBS
+        };
+        std::vector<double> significance_models_;
+        Xaode() : nFeatures_{ 0 }, statesClass_{ 0 }, matrixState_{ MatrixState::EMPTY } {}
+        // -------------------------------------------------------
+        // fit
+        // -------------------------------------------------------
+        //
+        // Classifiers interface
+        // all parameter decide if the model is initialized with all the parents active or none of them
+        //
+        // states.size() = nFeatures + 1,
+        //   where states.back() = number of class states.
+        //
+        // We'll store:
+        //  1) p(x_i=si | c) in classFeatureProbs_
+        //  2) p(x_j=sj | c, x_i=si) in data_, with i<j => i is "superparent," j is "child."
+        //
+        // Internally, in COUNTS mode, data_ accumulates raw counts, then
+        // computeProbabilities(...) normalizes them into conditionals.
+        void fit(std::vector<std::vector<int>>& X, std::vector<int>& y, const std::vector<std::string>& features, const std::string& className, std::map<std::string, std::vector<int>>& states, const torch::Tensor& weights, const bool all_parents, const bayesnet::Smoothing_t smoothing)
+        {
+            int num_instances = X[0].size();
+            nFeatures_ = X.size();
+
+            significance_models_.resize(nFeatures_, (all_parents ? 1.0 : 0.0));
+            for (int i = 0; i < nFeatures_; i++) {
+                if (all_parents) active_parents.push_back(i);
+                states_.push_back(*max_element(X[i].begin(), X[i].end()) + 1);
+            }
+            states_.push_back(*max_element(y.begin(), y.end()) + 1);
+            //
+            statesClass_ = states_.back();
+            classCounts_.resize(statesClass_, 0.0);
+            classPriors_.resize(statesClass_, 0.0);
+            //
+            // Initialize data structures
+            //
+            active_parents.resize(nFeatures_);
+            int totalStates = std::accumulate(states_.begin(), states_.end(), 0) - statesClass_;
+
+            // For p(x_i=si | c), we store them in a 1D array classFeatureProbs_ after we compute.
+            // We'll need the offsets for each feature i in featureClassOffset_.
+            featureClassOffset_.resize(nFeatures_);
+            // We'll store p(x_child=sj | c, x_sp=si) for each pair (i<j).
+            // So data_(i, si, j, sj, c) indexes into a big 1D array with an offset.
+            // For p(x_i=si | c), we store them in a 1D array classFeatureProbs_ after we compute.
+            // We'll need the offsets for each feature i in featureClassOffset_.
+            featureClassOffset_.resize(nFeatures_);
+            pairOffset_.resize(totalStates);
+            int feature_offset = 0;
+            int runningOffset = 0;
+            int feature = 0, index = 0;
+            for (int i = 0; i < nFeatures_; ++i) {
+                featureClassOffset_[i] = feature_offset;
+                feature_offset += states_[i];
+                for (int j = 0; j < states_[i]; ++j) {
+                    pairOffset_[feature++] = index;
+                    index += runningOffset;
+                }
+                runningOffset += states_[i];
+            }
+            int totalSize = index * statesClass_;
+            data_.resize(totalSize);
+            dataOpp_.resize(totalSize);
+
+            classFeatureCounts_.resize(feature_offset * statesClass_);
+            classFeatureProbs_.resize(feature_offset * statesClass_);
+
+            matrixState_ = MatrixState::COUNTS;
+            //
+            // Add samples
+            //
+            std::vector<int> instance(nFeatures_ + 1);
+            for (int n_instance = 0; n_instance < num_instances; n_instance++) {
+                for (int feature = 0; feature < nFeatures_; feature++) {
+                    instance[feature] = X[feature][n_instance];
+                }
+                instance[nFeatures_] = y[n_instance];
+                addSample(instance, weights[n_instance].item<double>());
+            }
+            switch (smoothing) {
+                case bayesnet::Smoothing_t::ORIGINAL:
+                    alpha_ = 1.0 / num_instances;
+                    break;
+                case bayesnet::Smoothing_t::LAPLACE:
+                    alpha_ = 1.0;
+                    break;
+                default:
+                    alpha_ = 0.0; // No smoothing 
+            }
+            initializer_ = std::numeric_limits<double>::max() / (nFeatures_ * nFeatures_);
+            computeProbabilities();
+        }
+        std::string to_string() const
+        {
+            std::ostringstream ostream;
+            ostream << "-------- Xaode.status --------" << std::endl
+                << "- nFeatures = " << nFeatures_ << std::endl
+                << "- statesClass = " << statesClass_ << std::endl
+                << "- matrixState = " << (matrixState_ == MatrixState::COUNTS ? "COUNTS" : "PROBS") << std::endl;
+            ostream << "- states: size: " << states_.size() << std::endl;
+            for (int s : states_) ostream << s << " "; ostream << std::endl;
+            ostream << "- classCounts: size: " << classCounts_.size() << std::endl;
+            for (double cc : classCounts_) ostream << cc << " "; ostream << std::endl;
+            ostream << "- classPriors: size: " << classPriors_.size() << std::endl;
+            for (double cp : classPriors_) ostream << cp << " "; ostream << std::endl;
+            ostream << "- classFeatureCounts: size: " << classFeatureCounts_.size() << std::endl;
+            for (double cfc : classFeatureCounts_) ostream << cfc << " "; ostream << std::endl;
+            ostream << "- classFeatureProbs: size: " << classFeatureProbs_.size() << std::endl;
+            for (double cfp : classFeatureProbs_) ostream << cfp << " "; ostream << std::endl;
+            ostream << "- featureClassOffset: size: " << featureClassOffset_.size() << std::endl;
+            for (int f : featureClassOffset_) ostream << f << " "; ostream << std::endl;
+            ostream << "- pairOffset_: size: " << pairOffset_.size() << std::endl;
+            for (int p : pairOffset_) ostream << p << " "; ostream << std::endl;
+            ostream << "- data: size: " << data_.size() << std::endl;
+            for (double d : data_) ostream << d << " "; ostream << std::endl;
+            ostream << "- dataOpp: size: " << dataOpp_.size() << std::endl;
+            for (double d : dataOpp_) ostream << d << " "; ostream << std::endl;
+            ostream << "--------------------------------" << std::endl;
+            std::string output = ostream.str();
+            return output;
+        }
+        // -------------------------------------------------------
+        // addSample (only in COUNTS mode)
+        // -------------------------------------------------------
+        // 
+        // instance should have the class at the end.
+        // 
+        void addSample(const std::vector<int>& instance, double weight)
+        {
+            //
+            // (A) increment classCounts_
+            // (B) increment feature–class counts => for p(x_i|c)
+            // (C) increment pair (superparent= i, child= j) counts => data_ 
+            //
+            int c = instance.back();
+            if (weight <= 0.0) {
+                return;
+            }
+            // (A) increment classCounts_
+            classCounts_[c] += weight;
+
+            // (B,C)
+            // We'll store raw counts now and turn them into p(child| c, superparent) later.
+            int idx, fcIndex, sp, sc, i_offset;
+            for (int parent = 0; parent < nFeatures_; ++parent) {
+                sp = instance[parent];
+                // (B) increment feature–class counts => for p(x_i|c)
+                fcIndex = (featureClassOffset_[parent] + sp) * statesClass_ + c;
+                classFeatureCounts_[fcIndex] += weight;
+                // (C) increment pair (superparent= i, child= j) counts => data_
+                i_offset = pairOffset_[featureClassOffset_[parent] + sp];
+                for (int child = 0; child < parent; ++child) {
+                    sc = instance[child];
+                    idx = (i_offset + featureClassOffset_[child] + sc) * statesClass_ + c;
+                    data_[idx] += weight;
+                }
+            }
+        }
+        // -------------------------------------------------------
+        // computeProbabilities
+        // -------------------------------------------------------
+        //
+        // Once all samples are added in COUNTS mode, call this to:
+        //  1) compute p(c) => classPriors_
+        //  2) compute p(x_i=si | c) => classFeatureProbs_
+        //  3) compute p(x_j=sj | c, x_i=si) => data_ (for i<j) dataOpp_ (for i>j)
+        //
+        void computeProbabilities()
+        {
+            if (matrixState_ != MatrixState::COUNTS) {
+                throw std::logic_error("computeProbabilities: must be in COUNTS mode.");
+            }
+            double totalCount = std::accumulate(classCounts_.begin(), classCounts_.end(), 0.0);
+            // (1) p(c)
+            if (totalCount <= 0.0) {
+                // fallback => uniform
+                double unif = 1.0 / statesClass_;
+                for (int c = 0; c < statesClass_; ++c) {
+                    classPriors_[c] = unif;
+                }
+            } else {
+                for (int c = 0; c < statesClass_; ++c) {
+                    classPriors_[c] = (classCounts_[c] + alpha_) / (totalCount + alpha_ * statesClass_);
+                }
+            }
+            // (2) p(x_i=si | c) => classFeatureProbs_
+            int idx, sf;
+            double denom;
+            for (int feature = 0; feature < nFeatures_; ++feature) {
+                sf = states_[feature];
+                for (int c = 0; c < statesClass_; ++c) {
+                    denom = classCounts_[c] + alpha_ * sf;
+                    for (int sf_value = 0; sf_value < sf; ++sf_value) {
+                        idx = (featureClassOffset_[feature] + sf_value) * statesClass_ + c;
+                        classFeatureProbs_[idx] = (classFeatureCounts_[idx] + alpha_) / denom;
+                    }
+                }
+            }
+            // getCountFromTable(int classVal, int pIndex, int childIndex)
+            // (3) p(x_c=sc | c, x_p=sp) => data_(parent,sp,child,sc,c)
+            // (3) p(x_p=sp | c, x_c=sc) => dataOpp_(child,sc,parent,sp,c)
+            //                    C(x_c, x_p, c) + alpha_
+            // P(x_p | x_c, c) = -----------------------------------
+            //                           C(x_c, c) + alpha_
+            double pcc_count, pc_count, cc_count;
+            double conditionalProb, oppositeCondProb;
+            int part1, part2, p1, part2_class, p1_class;
+            for (int parent = 1; parent < nFeatures_; ++parent) {
+                for (int sp = 0; sp < states_[parent]; ++sp) {
+                    p1 = featureClassOffset_[parent] + sp;
+                    part1 = pairOffset_[p1];
+                    p1_class = p1 * statesClass_;
+                    for (int child = 0; child < parent; ++child) {
+                        for (int sc = 0; sc < states_[child]; ++sc) {
+                            part2 = featureClassOffset_[child] + sc;
+                            part2_class = part2 * statesClass_;
+                            for (int c = 0; c < statesClass_; c++) {
+                                idx = (part1 + part2) * statesClass_ + c;
+                                // Parent, Child, Class Count
+                                pcc_count = data_[idx];
+                                // Parent, Class count
+                                pc_count = classFeatureCounts_[p1_class + c];
+                                // Child, Class count
+                                cc_count = classFeatureCounts_[part2_class + c];
+                                // p(x_c=sc | c, x_p=sp)
+                                conditionalProb = (pcc_count + alpha_) / (pc_count + alpha_ * states_[child]);
+                                data_[idx] = conditionalProb;
+                                // p(x_p=sp | c, x_c=sc)
+                                oppositeCondProb = (pcc_count + alpha_) / (cc_count + alpha_ * states_[parent]);
+                                dataOpp_[idx] = oppositeCondProb;
+                            }
+                        }
+                    }
+                }
+            }
+            matrixState_ = MatrixState::PROBS;
+        }
+        // -------------------------------------------------------
+        // predict_proba_spode
+        // -------------------------------------------------------
+        //
+        // Single-superparent approach:
+        // P(c | x) ∝ p(c) * p(x_sp| c) * ∏_{i≠sp} p(x_i | c, x_sp)
+        //
+        // 'instance' should have size == nFeatures_ (no class).
+        // sp in [0..nFeatures_).
+        // We multiply p(c) * p(x_sp| c) * p(x_i| c, x_sp).
+        // Then normalize the distribution.
+        //
+        std::vector<double> predict_proba_spode(const std::vector<int>& instance, int parent)
+        {
+            // accumulates posterior probabilities for each class
+            auto probs = std::vector<double>(statesClass_);
+            auto spodeProbs = std::vector<double>(statesClass_, 0.0);
+            if (std::find(active_parents.begin(), active_parents.end(), parent) == active_parents.end()) {
+                return spodeProbs;
+            }
+            // Initialize the probabilities with the feature|class probabilities x class priors
+            int localOffset;
+            int sp = instance[parent];
+            localOffset = (featureClassOffset_[parent] + sp) * statesClass_;
+            for (int c = 0; c < statesClass_; ++c) {
+                spodeProbs[c] = classFeatureProbs_[localOffset + c] * classPriors_[c] * initializer_;
+            }
+            int idx, base, sc, parent_offset;
+            for (int child = 0; child < nFeatures_; ++child) {
+                if (child == parent) {
+                    continue;
+                }
+                sc = instance[child];
+                if (child > parent) {
+                    parent_offset = pairOffset_[featureClassOffset_[child] + sc];
+                    base = (parent_offset + featureClassOffset_[parent] + sp) * statesClass_;
+                } else {
+                    parent_offset = pairOffset_[featureClassOffset_[parent] + sp];
+                    base = (parent_offset + featureClassOffset_[child] + sc) * statesClass_;
+                }
+                for (int c = 0; c < statesClass_; ++c) {
+                    /*
+                    * The probability P(xc|xp,c) is stored in dataOpp_, and
+                    * the probability P(xp|xc,c) is stored in data_
+                    */
+                    idx = base + c;
+                    double factor = child > parent ? dataOpp_[idx] : data_[idx];
+                    // double factor = data_[idx];
+                    spodeProbs[c] *= factor;
+                }
+            }
+            // Normalize the probabilities
+            normalize(spodeProbs);
+            return spodeProbs;
+        }
+        int predict_spode(const std::vector<int>& instance, int parent)
+        {
+            auto probs = predict_proba_spode(instance, parent);
+            return (int)std::distance(probs.begin(), std::max_element(probs.begin(), probs.end()));
+        }
+        // -------------------------------------------------------
+        // predict_proba
+        // -------------------------------------------------------
+        //
+        // P(c | x) ∝ p(c) * ∏_{i} p(x_i | c) * ∏_{i<j} p(x_j | c, x_i) * p(x_i | c, x_j)
+        //
+        // 'instance' should have size == nFeatures_ (no class).
+        // We multiply p(c) * p(x_i| c) * p(x_j| c, x_i) for all i, j.
+        // Then normalize the distribution.
+        //
+        std::vector<double> predict_proba(const std::vector<int>& instance)
+        {
+            // accumulates posterior probabilities for each class
+            auto probs = std::vector<double>(statesClass_);
+            auto spodeProbs = std::vector<std::vector<double>>(nFeatures_, std::vector<double>(statesClass_));
+            // Initialize the probabilities with the feature|class probabilities
+            int localOffset;
+            for (int feature = 0; feature < nFeatures_; ++feature) {
+                // if feature is not in the active_parents, skip it
+                if (std::find(active_parents.begin(), active_parents.end(), feature) == active_parents.end()) {
+                    continue;
+                }
+                localOffset = (featureClassOffset_[feature] + instance[feature]) * statesClass_;
+                for (int c = 0; c < statesClass_; ++c) {
+                    spodeProbs[feature][c] = classFeatureProbs_[localOffset + c] * classPriors_[c] * initializer_;
+                }
+            }
+            int idx, base, sp, sc, parent_offset;
+            for (int parent = 1; parent < nFeatures_; ++parent) {
+                // if parent is not in the active_parents, skip it
+                if (std::find(active_parents.begin(), active_parents.end(), parent) == active_parents.end()) {
+                    continue;
+                }
+                sp = instance[parent];
+                parent_offset = pairOffset_[featureClassOffset_[parent] + sp];
+                for (int child = 0; child < parent; ++child) {
+                    sc = instance[child];
+                    if (child > parent) {
+                        parent_offset = pairOffset_[featureClassOffset_[child] + sc];
+                        base = (parent_offset + featureClassOffset_[parent] + sp) * statesClass_;
+                    } else {
+                        parent_offset = pairOffset_[featureClassOffset_[parent] + sp];
+                        base = (parent_offset + featureClassOffset_[child] + sc) * statesClass_;
+                    }
+                    for (int c = 0; c < statesClass_; ++c) {
+                        /*
+                         * The probability P(xc|xp,c) is stored in dataOpp_, and
+                         * the probability P(xp|xc,c) is stored in data_
+                         */
+                        idx = base + c;
+                        double factor_child = child > parent ? data_[idx] : dataOpp_[idx];
+                        double factor_parent = child > parent ? dataOpp_[idx] : data_[idx];
+                        spodeProbs[child][c] *= factor_child;
+                        spodeProbs[parent][c] *= factor_parent;
+                    }
+                }
+            }
+            /* add all the probabilities for each class */
+            for (int c = 0; c < statesClass_; ++c) {
+                for (int i = 0; i < nFeatures_; ++i) {
+                    probs[c] += spodeProbs[i][c] * significance_models_[i];
+                }
+            }
+            // Normalize the probabilities
+            normalize(probs);
+            return probs;
+        }
+        void normalize(std::vector<double>& probs) const
+        {
+            double sum = std::accumulate(probs.begin(), probs.end(), 0.0);
+            if (std::isnan(sum)) {
+                throw std::runtime_error("Can't normalize array. Sum is NaN.");
+            }
+            if (sum == 0) {
+                return;
+            }
+            for (int i = 0; i < (int)probs.size(); i++) {
+                probs[i] /= sum;
+            }
+        }
+        // Returns current mode: INIT, COUNTS or PROBS
+        MatrixState state() const
+        {
+            return matrixState_;
+        }
+        int statesClass() const
+        {
+            return statesClass_;
+        }
+        int nFeatures() const
+        {
+            return nFeatures_;
+        }
+        int getNumberOfStates() const
+        {
+            return std::accumulate(states_.begin(), states_.end(), 0) * nFeatures_;
+        }
+        int getNumberOfEdges() const
+        {
+            return nFeatures_ * (2 * nFeatures_ - 1);
+        }
+        int getNumberOfNodes() const
+        {
+            return (nFeatures_ + 1) * nFeatures_;
+        }
+        void add_active_parent(int active_parent)
+        {
+            active_parents.push_back(active_parent);
+        }
+        void remove_last_parent()
+        {
+            active_parents.pop_back();
+        }
+
+    private:
+        // -----------
+        // MEMBER DATA
+        // -----------
+        std::vector<int> states_;            // [states_feat0, ..., states_feat(n-1), statesClass_]
+        int nFeatures_;
+        int statesClass_;
+
+        // data_ means p(child=sj | c, superparent= si) after normalization.
+        // But in COUNTS mode, it accumulates raw counts.
+        std::vector<int> pairOffset_;
+        // data_ stores p(child=sj | c, superparent=si) for each pair (i<j).
+        std::vector<double> data_;
+        // dataOpp_ stores p(superparent=si | c, child=sj) for each pair (i<j).
+        std::vector<double> dataOpp_;
+
+        // classCounts_[c]
+        std::vector<double> classCounts_;
+        std::vector<double> classPriors_;       // => p(c)
+
+        // For p(x_i=si| c), we store counts in classFeatureCounts_ => offset by featureClassOffset_[i]
+        std::vector<int> featureClassOffset_;
+        std::vector<double> classFeatureCounts_;
+        std::vector<double> classFeatureProbs_;  // => p(x_i=si | c) after normalization
+
+        MatrixState matrixState_;
+
+        double alpha_ = 1.0; // Laplace smoothing
+        double initializer_ = 1.0;
+        std::vector<int> active_parents;
+    };
+}
+#endif // XAODE_H

src/grid/GridBase.cpp

@@ -0,0 +1,314 @@
+#include <random>
+#include <cstddef>
+#include "common/DotEnv.h"
+#include "common/Paths.h"
+#include "common/Colors.h"
+#include "GridBase.h"
+
+
+namespace platform {
+
+    GridBase::GridBase(struct ConfigGrid& config)
+    {
+        this->config = config;
+        auto env = platform::DotEnv();
+        this->config.platform = env.get("platform");
+
+    }
+    void GridBase::validate_config()
+    {
+        if (config.smooth_strategy == "ORIGINAL")
+            smooth_type = bayesnet::Smoothing_t::ORIGINAL;
+        else if (config.smooth_strategy == "LAPLACE")
+            smooth_type = bayesnet::Smoothing_t::LAPLACE;
+        else if (config.smooth_strategy == "CESTNIK")
+            smooth_type = bayesnet::Smoothing_t::CESTNIK;
+        else {
+            std::cerr << "GridBase: Unknown smoothing strategy: " << config.smooth_strategy << std::endl;
+            exit(1);
+        }
+    }
+    std::string GridBase::get_color_rank(int rank)
+    {
+        auto colors = { Colors::WHITE(), Colors::RED(), Colors::GREEN(),  Colors::BLUE(), Colors::MAGENTA(), Colors::CYAN(), Colors::YELLOW(), Colors::BLACK() };
+        std::string id = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
+        auto idx = rank % id.size();
+        return *(colors.begin() + rank % colors.size()) + id[idx];
+    }
+    void GridBase::shuffle_and_progress_bar(json& tasks)
+    {
+        // Shuffle the array so heavy datasets are eas  ier spread across the workers
+        std::mt19937 g{ 271 }; // Use fixed seed to obtain the same shuffle
+        std::shuffle(tasks.begin(), tasks.end(), g);
+        std::cout << "* Number of tasks: " << tasks.size() << std::endl;
+        std::cout << separator << std::flush;
+        for (int i = 0; i < tasks.size(); ++i) {
+            if ((i + 1) % 10 == 0)
+                std::cout << separator;
+            else
+                std::cout << (i + 1) % 10;
+        }
+        std::cout << separator << std::endl << separator << std::flush;
+    }
+    json GridBase::build_tasks(Datasets& datasets)
+    {
+        /*
+        * Each task is a json object with the following structure:
+        * {
+        *   "dataset": "dataset_name",
+        *   "idx_dataset": idx_dataset, // used to identify the dataset in the results
+        *    // this index is relative to the list of used datasets in the actual run not to the whole datasets list
+        *   "seed": # of seed to use,
+        *   "fold": # of fold to process
+        * }
+        * This way a task consists in process all combinations of hyperparameters for a dataset, seed and fold
+        */
+        auto tasks = json::array();
+        auto all_datasets = datasets.getNames();
+        auto datasets_names = filterDatasets(datasets);
+        for (int idx_dataset = 0; idx_dataset < datasets_names.size(); ++idx_dataset) {
+            auto dataset = datasets_names[idx_dataset];
+            for (const auto& seed : config.seeds) {
+                for (int n_fold = 0; n_fold < config.n_folds; n_fold++) {
+                    json task = {
+                        { "dataset", dataset },
+                        { "idx_dataset", idx_dataset},
+                        { "seed", seed },
+                        { "fold", n_fold},
+                    };
+                    tasks.push_back(task);
+                }
+            }
+        }
+        shuffle_and_progress_bar(tasks);
+        return tasks;
+    }
+    void GridBase::summary(json& all_results, json& tasks, struct ConfigMPI& config_mpi)
+    {
+        // Report the tasks done by each worker, showing dataset number, seed, fold and time spent
+        // The format I want to show is:
+        // worker, dataset, seed, fold, time
+        // with headers
+        std::cout << Colors::RESET() << "* Summary of tasks done by each worker" << std::endl;
+        json worker_tasks = json::array();
+        for (int i = 0; i < config_mpi.n_procs; ++i) {
+            worker_tasks.push_back(json::array());
+        }
+        int max_dataset = 7;
+        for (const auto& [key, results] : all_results.items()) {
+            auto dataset = key;
+            if (dataset.size() > max_dataset)
+                max_dataset = dataset.size();
+            for (const auto& result : results) {
+                int n_task = result["task"].get<int>();
+                json task = tasks[n_task];
+                auto seed = task["seed"].get<int>();
+                auto fold = task["fold"].get<int>();
+                auto time = result["time"].get<double>();
+                auto worker = result["process"].get<int>();
+                json line = {
+                    { "dataset", dataset },
+                    { "seed", seed },
+                    { "fold", fold },
+                    { "time", time }
+                };
+                worker_tasks[worker].push_back(line);
+            }
+        }
+        std::cout << Colors::MAGENTA() << " W  " << setw(max_dataset) << std::left << "Dataset";
+        std::cout << " Seed Fold Time" << std::endl;
+        std::cout << "=== " << std::string(max_dataset, '=') << " ==== ==== " << std::string(15, '=') << std::endl;
+        for (int worker = 0; worker < config_mpi.n_procs; ++worker) {
+            auto color = (worker % 2) ? Colors::CYAN() : Colors::BLUE();
+            std::cout << color << std::right << setw(3) << worker << " ";
+            if (worker == config_mpi.manager) {
+                std::cout << "Manager" << std::endl;
+                continue;
+            }
+            if (worker_tasks[worker].empty()) {
+                std::cout << "No tasks" << std::endl;
+                continue;
+            }
+            bool first = true;
+            double total = 0.0;
+            int num_tasks = 0;
+            for (const auto& task : worker_tasks[worker]) {
+                num_tasks++;
+                if (!first)
+                    std::cout << std::string(4, ' ');
+                else
+                    first = false;
+                std::cout << std::left << setw(max_dataset) << task["dataset"].get<std::string>();
+                std::cout << " " << setw(4) << std::right << task["seed"].get<int>();
+                std::cout << " " << setw(4) << task["fold"].get<int>();
+                std::cout << " " << setw(15) << std::setprecision(7) << std::fixed << task["time"].get<double>() << std::endl;
+                total += task["time"].get<double>();
+            }
+            if (num_tasks > 1) {
+                std::cout << Colors::MAGENTA() << "    ";
+                std::cout << setw(max_dataset) << "Total (" << setw(2) << std::right << num_tasks << ")" << std::string(7, '.');
+                std::cout << " " << setw(15) << std::setprecision(7) << std::fixed << total << std::endl;
+            }
+        }
+    }
+    void GridBase::go(struct ConfigMPI& config_mpi)
+    {
+        /*
+        * Each task is a json object with the data needed by the process
+        *
+        * The overall process consists in these steps:
+           * 0. Validate config, create the MPI result type & tasks
+           * 0.1 Create the MPI result type
+           * 0.2 Manager creates the tasks
+           * 1. Manager will broadcast the tasks to all the processes
+           * 1.1 Broadcast the number of tasks
+           * 1.2 Broadcast the length of the following string
+           * 1.2 Broadcast the tasks as a char* string
+           * 2a. Producer delivers the tasks to the consumers
+           * 2a.1 Producer will loop to send all the tasks to the consumers and receive the results
+           * 2a.2 Producer will send the end message to all the consumers
+           * 2b. Consumers process the tasks and send the results to the producer
+           * 2b.1 Consumers announce to the producer that they are ready to receive a task
+           * 2b.2 Consumers receive the task from the producer and process it
+           * 2b.3 Consumers send the result to the producer
+           * 3. Manager compile results for each dataset
+           * 3.1 Loop thru all the results obtained from each outer fold (task) and select the best
+           * 3.2 Save the results
+           * 3.3 Summary of jobs done
+        */
+        //
+        // 0.1 Create the MPI result type
+        //
+        validate_config();
+        Task_Result result;
+        int tasks_size;
+        MPI_Datatype MPI_Result;
+        MPI_Datatype type[11] = { MPI_UNSIGNED, MPI_UNSIGNED, MPI_INT, MPI_DOUBLE, MPI_DOUBLE, MPI_DOUBLE, MPI_DOUBLE, MPI_DOUBLE, MPI_DOUBLE, MPI_INT, MPI_INT };
+        int blocklen[11] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 };
+        MPI_Aint disp[11];
+        disp[0] = offsetof(Task_Result, idx_dataset);
+        disp[1] = offsetof(Task_Result, idx_combination);
+        disp[2] = offsetof(Task_Result, n_fold);
+        disp[3] = offsetof(Task_Result, score);
+        disp[4] = offsetof(Task_Result, time);
+        disp[5] = offsetof(Task_Result, time_train);
+        disp[6] = offsetof(Task_Result, nodes);
+        disp[7] = offsetof(Task_Result, leaves);
+        disp[8] = offsetof(Task_Result, depth);
+        disp[9] = offsetof(Task_Result, process);
+        disp[10] = offsetof(Task_Result, task);
+        MPI_Type_create_struct(11, blocklen, disp, type, &MPI_Result);
+        MPI_Type_commit(&MPI_Result);
+        //
+        // 0.2 Manager creates the tasks
+        //
+        char* msg;
+        json tasks;
+        auto env = platform::DotEnv();
+        auto datasets = Datasets(config.discretize, Paths::datasets(), env.get("discretize_algo"));
+        if (config_mpi.rank == config_mpi.manager) {
+            timer.start();
+            tasks = build_tasks(datasets);
+            auto tasks_str = tasks.dump();
+            tasks_size = tasks_str.size();
+            msg = new char[tasks_size + 1];
+            strcpy(msg, tasks_str.c_str());
+        }
+        //
+        // 1. Manager will broadcast the tasks to all the processes
+        //
+        MPI_Bcast(&tasks_size, 1, MPI_INT, config_mpi.manager, MPI_COMM_WORLD);
+        if (config_mpi.rank != config_mpi.manager) {
+            msg = new char[tasks_size + 1];
+        }
+        MPI_Bcast(msg, tasks_size + 1, MPI_CHAR, config_mpi.manager, MPI_COMM_WORLD);
+        tasks = json::parse(msg);
+        delete[] msg;
+
+
+        if (config_mpi.rank == config_mpi.manager) {
+            //
+            // 2a. Producer delivers the tasks to the consumers
+            //
+            auto datasets_names = filterDatasets(datasets);
+            json all_results = producer(datasets_names, tasks, config_mpi, MPI_Result);
+            std::cout << separator << std::endl;
+            //
+            // 3. Manager compile results for each dataset
+            //
+            auto results = initializeResults();
+            compile_results(results, all_results, config.model);
+            //
+            // 3.2 Save the results
+            //
+            save(results);
+            //
+            // 3.3 Summary of jobs done
+            //
+            if (!config.quiet)
+                summary(all_results, tasks, config_mpi);
+        } else {
+            //
+            // 2b. Consumers process the tasks and send the results to the producer
+            //
+            consumer(datasets, tasks, config, config_mpi, MPI_Result);
+        }
+    }
+    json GridBase::producer(std::vector<std::string>& names, json& tasks, struct ConfigMPI& config_mpi, MPI_Datatype& MPI_Result)
+    {
+        Task_Result result;
+        json results;
+        int num_tasks = tasks.size();
+        //
+        // 2a.1 Producer will loop to send all the tasks to the consumers and receive the results
+        //
+        for (int i = 0; i < num_tasks; ++i) {
+            MPI_Status status;
+            MPI_Recv(&result, 1, MPI_Result, MPI_ANY_SOURCE, MPI_ANY_TAG, MPI_COMM_WORLD, &status);
+            if (status.MPI_TAG == TAG_RESULT) {
+                //Store result
+                store_result(names, result, results);
+
+            }
+            MPI_Send(&i, 1, MPI_INT, status.MPI_SOURCE, TAG_TASK, MPI_COMM_WORLD);
+        }
+        //
+        // 2a.2 Producer will send the end message to all the consumers
+        //
+        for (int i = 0; i < config_mpi.n_procs - 1; ++i) {
+            MPI_Status status;
+            MPI_Recv(&result, 1, MPI_Result, MPI_ANY_SOURCE, MPI_ANY_TAG, MPI_COMM_WORLD, &status);
+            if (status.MPI_TAG == TAG_RESULT) {
+                //Store result
+                store_result(names, result, results);
+            }
+            MPI_Send(&i, 1, MPI_INT, status.MPI_SOURCE, TAG_END, MPI_COMM_WORLD);
+        }
+        return results;
+    }
+    void GridBase::consumer(Datasets& datasets, json& tasks, struct ConfigGrid& config, struct ConfigMPI& config_mpi, MPI_Datatype& MPI_Result)
+    {
+        Task_Result result;
+        //
+        // 2b.1 Consumers announce to the producer that they are ready to receive a task
+        //
+        MPI_Send(&result, 1, MPI_Result, config_mpi.manager, TAG_QUERY, MPI_COMM_WORLD);
+        int task;
+        while (true) {
+            MPI_Status status;
+            //
+            // 2b.2 Consumers receive the task from the producer and process it
+            //
+            MPI_Recv(&task, 1, MPI_INT, config_mpi.manager, MPI_ANY_TAG, MPI_COMM_WORLD, &status);
+            if (status.MPI_TAG == TAG_END) {
+                break;
+            }
+            consumer_go(config, config_mpi, tasks, task, datasets, &result);
+            //
+            // 2b.3 Consumers send the result to the producer
+            //
+            MPI_Send(&result, 1, MPI_Result, config_mpi.manager, TAG_RESULT, MPI_COMM_WORLD);
+        }
+    }
+
+}

src/grid/GridBase.h

@@ -0,0 +1,39 @@
+#ifndef GRIDBASE_H
+#define GRIDBASE_H
+#include <string>
+#include <mpi.h>
+#include <nlohmann/json.hpp>
+#include "common/Datasets.h"
+#include "common/Timer.hpp"
+#include "main/HyperParameters.h"
+#include "GridConfig.h"
+
+
+namespace platform {
+    using json = nlohmann::ordered_json;
+    class GridBase {
+    public:
+        explicit GridBase(struct ConfigGrid& config);
+        ~GridBase() = default;
+        void go(struct ConfigMPI& config_mpi);
+        void validate_config();
+    protected:
+        json build_tasks(Datasets& datasets);
+        virtual void save(json& results) = 0;
+        virtual std::vector<std::string> filterDatasets(Datasets& datasets) const = 0;
+        virtual json initializeResults() = 0;
+        virtual void compile_results(json& results, json& all_results, std::string& model) = 0;
+        virtual json store_result(std::vector<std::string>& names, Task_Result& result, json& results) = 0;
+        virtual void consumer_go(struct ConfigGrid& config, struct ConfigMPI& config_mpi, json& tasks, int n_task, Datasets& datasets, Task_Result* result) = 0;
+        void shuffle_and_progress_bar(json& tasks);
+        json producer(std::vector<std::string>& names, json& tasks, struct ConfigMPI& config_mpi, MPI_Datatype& MPI_Result);
+        void consumer(Datasets& datasets, json& tasks, struct ConfigGrid& config, struct ConfigMPI& config_mpi, MPI_Datatype& MPI_Result);
+        std::string get_color_rank(int rank);
+        void summary(json& all_results, json& tasks, struct ConfigMPI& config_mpi);
+        struct ConfigGrid config;
+        Timer timer; // used to measure the time of the whole process
+        const std::string separator = "|";
+        bayesnet::Smoothing_t smooth_type{ bayesnet::Smoothing_t::NONE };
+    };
+} /* namespace platform */
+#endif

src/grid/GridConfig.h

@@ -0,0 +1,55 @@
+#ifndef GRIDCONFIG_H
+#define GRIDCONFIG_H
+#include <string>
+#include <map>
+#include <mpi.h>
+#include <nlohmann/json.hpp>
+#include "common/Datasets.h"
+#include "common/Timer.hpp"
+#include "main/HyperParameters.h"
+#include "GridData.h"
+#include "GridConfig.h"
+#include "bayesnet/network/Network.h"
+
+
+namespace platform {
+    using json = nlohmann::ordered_json;
+    struct ConfigGrid {
+        std::string model;
+        std::string score;
+        std::string continue_from;
+        std::string platform;
+        std::string smooth_strategy;
+        bool quiet;
+        bool only; // used with continue_from to only compute that dataset
+        bool discretize;
+        bool stratified;
+        int nested;
+        int n_folds;
+        json excluded;
+        std::vector<int> seeds;
+    };
+    struct ConfigMPI {
+        int rank;
+        int n_procs;
+        int manager;
+    };
+    typedef struct {
+        uint idx_dataset;
+        uint idx_combination;
+        int n_fold;
+        double score; // Experiment: Score test, no score train in this case
+        double time; // Experiment: Time test
+        double time_train;
+        double nodes; // Experiment specific
+        double leaves; // Experiment specific
+        double depth; // Experiment specific
+        int process;
+        int task;
+    } Task_Result;
+    const int TAG_QUERY = 1;
+    const int TAG_RESULT = 2;
+    const int TAG_TASK = 3;
+    const int TAG_END = 4;
+} /* namespace platform */
+#endif

src/grid/GridExperiment.cpp

@@ -0,0 +1,196 @@
+#include <iostream>
+#include <cstddef>
+#include <torch/torch.h>
+#include <folding.hpp>
+#include "main/Models.h"
+#include "common/Paths.h"
+#include "common/Utils.h"
+#include "GridExperiment.h"
+
+namespace platform {
+    // GridExperiment::GridExperiment(argparse::ArgumentParser& program, struct ConfigGrid& config) : arguments(program), GridBase(config)
+    GridExperiment::GridExperiment(ArgumentsExperiment& program, struct ConfigGrid& config) : arguments(program), GridBase(config)
+    {
+        experiment = arguments.initializedExperiment();
+        filesToTest = arguments.getFilesToTest();
+        saveResults = arguments.haveToSaveResults();
+        this->config.model = experiment.getModel();
+        this->config.score = experiment.getScore();
+        this->config.discretize = experiment.isDiscretized();
+        this->config.stratified = experiment.isStratified();
+        this->config.smooth_strategy = experiment.getSmoothStrategy();
+        this->config.n_folds = experiment.getNFolds();
+        this->config.seeds = experiment.getRandomSeeds();
+        this->config.quiet = experiment.isQuiet();
+    }
+    json GridExperiment::getResults()
+    {
+        return computed_results;
+    }
+    std::vector<std::string> GridExperiment::filterDatasets(Datasets& datasets) const
+    {
+        return filesToTest;
+    }
+    json GridExperiment::initializeResults()
+    {
+        json results;
+        return results;
+    }
+    void GridExperiment::save(json& results)
+    {
+    }
+    void GridExperiment::compile_results(json& results, json& all_results, std::string& model)
+    {
+        auto datasets = Datasets(false, Paths::datasets());
+        nlohmann::json temp = all_results; // To restore the order of the data by dataset name
+        all_results = temp;
+        for (const auto& result_item : all_results.items()) {
+            // each result has the results of all the outer folds as each one were a different task
+            auto dataset_name = result_item.key();
+            auto data = result_item.value();
+            auto result = json::object();
+            int data_size = data.size();
+            auto score = torch::zeros({ data_size }, torch::kFloat64);
+            auto score_train = torch::zeros({ data_size }, torch::kFloat64);
+            auto time_test = torch::zeros({ data_size }, torch::kFloat64);
+            auto time_train = torch::zeros({ data_size }, torch::kFloat64);
+            auto nodes = torch::zeros({ data_size }, torch::kFloat64);
+            auto leaves = torch::zeros({ data_size }, torch::kFloat64);
+            auto depth = torch::zeros({ data_size }, torch::kFloat64);
+            auto& dataset = datasets.getDataset(dataset_name);
+            dataset.load();
+            //
+            // Prepare Result
+            //
+            auto partial_result = PartialResult();
+            partial_result.setSamples(dataset.getNSamples()).setFeatures(dataset.getNFeatures()).setClasses(dataset.getNClasses());
+            partial_result.setHyperparameters(experiment.getHyperParameters().get(dataset_name));
+            for (int fold = 0; fold < data_size; ++fold) {
+                partial_result.addScoreTest(data[fold]["score"]);
+                partial_result.addScoreTrain(0.0);
+                partial_result.addTimeTest(data[fold]["time"]);
+                partial_result.addTimeTrain(data[fold]["time_train"]);
+                score[fold] = data[fold]["score"].get<double>();
+                time_test[fold] = data[fold]["time"].get<double>();
+                time_train[fold] = data[fold]["time_train"].get<double>();
+                nodes[fold] = data[fold]["nodes"].get<double>();
+                leaves[fold] = data[fold]["leaves"].get<double>();
+                depth[fold] = data[fold]["depth"].get<double>();
+            }
+            partial_result.setGraph(std::vector<std::string>());
+            partial_result.setScoreTest(torch::mean(score).item<double>()).setScoreTrain(0.0);
+            partial_result.setScoreTestStd(torch::std(score).item<double>()).setScoreTrainStd(0.0);
+            partial_result.setTrainTime(torch::mean(time_train).item<double>()).setTestTime(torch::mean(time_test).item<double>());
+            partial_result.setTrainTimeStd(torch::std(time_train).item<double>()).setTestTimeStd(torch::std(time_test).item<double>());
+            partial_result.setNodes(torch::mean(nodes).item<double>()).setLeaves(torch::mean(leaves).item<double>()).setDepth(torch::mean(depth).item<double>());
+            partial_result.setDataset(dataset_name).setNotes(std::vector<std::string>());
+            partial_result.setConfusionMatrices(json::array());
+            experiment.addResult(partial_result);
+        }
+        auto clf = Models::instance()->create(experiment.getModel());
+        experiment.setModelVersion(clf->getVersion());
+        computed_results = results;
+    }
+    json GridExperiment::store_result(std::vector<std::string>& names, Task_Result& result, json& results)
+    {
+        json json_result = {
+            { "score", result.score },
+            { "combination", result.idx_combination },
+            { "fold", result.n_fold },
+            { "time", result.time },
+            { "time_train", result.time_train },
+            { "dataset", result.idx_dataset },
+            { "nodes", result.nodes },
+            { "leaves", result.leaves },
+            { "depth", result.depth },
+            { "process", result.process },
+            { "task", result.task }
+        };
+        auto name = names[result.idx_dataset];
+        if (!results.contains(name)) {
+            results[name] = json::array();
+        }
+        results[name].push_back(json_result);
+        return results;
+    }
+    void GridExperiment::consumer_go(struct ConfigGrid& config, struct ConfigMPI& config_mpi, json& tasks, int n_task, Datasets& datasets, Task_Result* result)
+    {
+        //
+        // initialize
+        //
+        Timer train_timer, test_timer;
+        json task = tasks[n_task];
+        auto model = config.model;
+        auto dataset_name = task["dataset"].get<std::string>();
+        auto idx_dataset = task["idx_dataset"].get<int>();
+        auto seed = task["seed"].get<int>();
+        auto n_fold = task["fold"].get<int>();
+        bool stratified = config.stratified;
+        bayesnet::Smoothing_t smooth;
+        if (config.smooth_strategy == "ORIGINAL")
+            smooth = bayesnet::Smoothing_t::ORIGINAL;
+        else if (config.smooth_strategy == "LAPLACE")
+            smooth = bayesnet::Smoothing_t::LAPLACE;
+        else if (config.smooth_strategy == "CESTNIK")
+            smooth = bayesnet::Smoothing_t::CESTNIK;
+        //
+        // Generate the hyperparameters combinations
+        //
+        auto& dataset = datasets.getDataset(dataset_name);
+        dataset.load();
+        auto [X, y] = dataset.getTensors();
+        auto features = dataset.getFeatures();
+        auto className = dataset.getClassName();
+        //
+        // Start working on task
+        //
+        folding::Fold* fold;
+        if (stratified)
+            fold = new folding::StratifiedKFold(config.n_folds, y, seed);
+        else
+            fold = new folding::KFold(config.n_folds, y.size(0), seed);
+        train_timer.start();
+        auto [train, test] = fold->getFold(n_fold);
+        auto [X_train, X_test, y_train, y_test] = dataset.getTrainTestTensors(train, test);
+        auto states = dataset.getStates(); // Get the states of the features Once they are discretized
+
+        //
+        // Build Classifier with selected hyperparameters
+        //
+        auto clf = Models::instance()->create(config.model);
+        auto valid = clf->getValidHyperparameters();
+        auto hyperparameters = experiment.getHyperParameters();
+        hyperparameters.check(valid, dataset_name);
+        clf->setHyperparameters(hyperparameters.get(dataset_name));
+        //
+        // Train model
+        //
+        clf->fit(X_train, y_train, features, className, states, smooth);
+        auto train_time = train_timer.getDuration();
+        //
+        // Test model
+        //
+        test_timer.start();
+        double score = clf->score(X_test, y_test);
+        delete fold;
+        auto test_time = test_timer.getDuration();
+        //
+        // Return the result
+        //
+        result->idx_dataset = task["idx_dataset"].get<int>();
+        result->idx_combination = 0;
+        result->score = score;
+        result->n_fold = n_fold;
+        result->time = test_time;
+        result->time_train = train_time;
+        result->nodes = clf->getNumberOfNodes();
+        result->leaves = clf->getNumberOfEdges();
+        result->depth = clf->getNumberOfStates();
+        result->process = config_mpi.rank;
+        result->task = n_task;
+        //
+        // Update progress bar
+        //
+        std::cout << get_color_rank(config_mpi.rank) << std::flush;
+    }
+} /* namespace platform */

src/grid/GridExperiment.h

@@ -0,0 +1,38 @@
+#ifndef GRIDEXPERIMENT_H
+#define GRIDEXPERIMENT_H
+#include <string>
+#include <mpi.h>
+#include <argparse/argparse.hpp>
+#include <nlohmann/json.hpp>
+#include "common/Datasets.h"
+#include "main/Experiment.h"
+#include "main/HyperParameters.h"
+#include "main/ArgumentsExperiment.h"
+#include "GridBase.h"
+
+
+namespace platform {
+    using json = nlohmann::ordered_json;
+    class GridExperiment : public GridBase {
+    public:
+        explicit GridExperiment(ArgumentsExperiment& program, struct ConfigGrid& config);
+        ~GridExperiment() = default;
+        json getResults();
+        Experiment& getExperiment() { return experiment; }
+        size_t numFiles() const { return filesToTest.size(); }
+        bool haveToSaveResults() const { return saveResults; }
+    private:
+        ArgumentsExperiment& arguments;
+        Experiment experiment;
+        json computed_results;
+        bool saveResults = false;
+        std::vector<std::string> filesToTest;
+        void save(json& results);
+        json initializeResults();
+        std::vector<std::string> filterDatasets(Datasets& datasets) const;
+        void compile_results(json& results, json& all_results, std::string& model);
+        json store_result(std::vector<std::string>& names, Task_Result& result, json& results);
+        void consumer_go(struct ConfigGrid& config, struct ConfigMPI& config_mpi, json& tasks, int n_task, Datasets& datasets, Task_Result* result);
+    };
+} /* namespace platform */
+#endif

src/grid/GridSearch.cpp

@@ -1,21 +1,14 @@
 #include <iostream>
-#include <cstddef>
 #include <torch/torch.h>
 #include <folding.hpp>
 #include "main/Models.h"
 #include "common/Paths.h"
-#include "common/Colors.h"
 #include "common/Utils.h"
+#include "common/Colors.h"
 #include "GridSearch.h"
 
 namespace platform {
-
-    std::string get_color_rank(int rank)
-    {
-        auto colors = { Colors::WHITE(), Colors::RED(), Colors::GREEN(),  Colors::BLUE(), Colors::MAGENTA(), Colors::CYAN() };
-        return *(colors.begin() + rank % colors.size());
-    }
-    GridSearch::GridSearch(struct ConfigGrid& config) : config(config)
+    GridSearch::GridSearch(struct ConfigGrid& config) : GridBase(config)
     {
     }
     json GridSearch::loadResults()
@@ -59,333 +52,13 @@ namespace platform {
         }
         return datasets_names;
     }
-    json GridSearch::build_tasks_mpi(int rank)
-    {
-        auto tasks = json::array();
-        auto grid = GridData(Paths::grid_input(config.model));
-        auto datasets = Datasets(false, Paths::datasets());
-        auto all_datasets = datasets.getNames();
-        auto datasets_names = filterDatasets(datasets);
-        for (int idx_dataset = 0; idx_dataset < datasets_names.size(); ++idx_dataset) {
-            auto dataset = datasets_names[idx_dataset];
-            for (const auto& seed : config.seeds) {
-                auto combinations = grid.getGrid(dataset);
-                for (int n_fold = 0; n_fold < config.n_folds; n_fold++) {
-                    json task = {
-                        { "dataset", dataset },
-                        { "idx_dataset", idx_dataset},
-                        { "seed", seed },
-                        { "fold", n_fold},
-                    };
-                    tasks.push_back(task);
-                }
-            }
-        }
-        // Shuffle the array so heavy datasets are spread across the workers
-        std::mt19937 g{ 271 }; // Use fixed seed to obtain the same shuffle
-        std::shuffle(tasks.begin(), tasks.end(), g);
-        std::cout << get_color_rank(rank) << "* Number of tasks: " << tasks.size() << std::endl;
-        std::cout << separator;
-        for (int i = 0; i < tasks.size(); ++i) {
-            std::cout << (i + 1) % 10;
-        }
-        std::cout << separator << std::endl << separator << std::flush;
-        return tasks;
-    }
-    void process_task_mpi_consumer(struct ConfigGrid& config, struct ConfigMPI& config_mpi, json& tasks, int n_task, Datasets& datasets, Task_Result* result)
-    {
-        // initialize
-        Timer timer;
-        timer.start();
-        json task = tasks[n_task];
-        auto model = config.model;
-        auto grid = GridData(Paths::grid_input(model));
-        auto dataset_name = task["dataset"].get<std::string>();
-        auto idx_dataset = task["idx_dataset"].get<int>();
-        auto seed = task["seed"].get<int>();
-        auto n_fold = task["fold"].get<int>();
-        bool stratified = config.stratified;
-        // Generate the hyperparamters combinations
-        auto& dataset = datasets.getDataset(dataset_name);
-        auto combinations = grid.getGrid(dataset_name);
-        dataset.load();
-        auto [X, y] = dataset.getTensors();
-        auto features = dataset.getFeatures();
-        auto className = dataset.getClassName();
-        //
-        // Start working on task
-        //
-        folding::Fold* fold;
-        if (stratified)
-            fold = new folding::StratifiedKFold(config.n_folds, y, seed);
-        else
-            fold = new folding::KFold(config.n_folds, y.size(0), seed);
-        auto [train, test] = fold->getFold(n_fold);
-        auto [X_train, X_test, y_train, y_test] = dataset.getTrainTestTensors(train, test);
-        auto states = dataset.getStates(); // Get the states of the features Once they are discretized
-        double best_fold_score = 0.0;
-        int best_idx_combination = -1;
-        bayesnet::Smoothing_t smoothing = bayesnet::Smoothing_t::NONE;
-        json best_fold_hyper;
-        for (int idx_combination = 0; idx_combination < combinations.size(); ++idx_combination) {
-            auto hyperparam_line = combinations[idx_combination];
-            auto hyperparameters = platform::HyperParameters(datasets.getNames(), hyperparam_line);
-            folding::Fold* nested_fold;
-            if (config.stratified)
-                nested_fold = new folding::StratifiedKFold(config.nested, y_train, seed);
-            else
-                nested_fold = new folding::KFold(config.nested, y_train.size(0), seed);
-            double score = 0.0;
-            for (int n_nested_fold = 0; n_nested_fold < config.nested; n_nested_fold++) {
-                // Nested level fold
-                auto [train_nested, test_nested] = nested_fold->getFold(n_nested_fold);
-                auto train_nested_t = torch::tensor(train_nested);
-                auto test_nested_t = torch::tensor(test_nested);
-                auto X_nested_train = X_train.index({ "...", train_nested_t });
-                auto y_nested_train = y_train.index({ train_nested_t });
-                auto X_nested_test = X_train.index({ "...", test_nested_t });
-                auto y_nested_test = y_train.index({ test_nested_t });
-                // Build Classifier with selected hyperparameters
-                auto clf = Models::instance()->create(config.model);
-                auto valid = clf->getValidHyperparameters();
-                hyperparameters.check(valid, dataset_name);
-                clf->setHyperparameters(hyperparameters.get(dataset_name));
-                // Train model
-                clf->fit(X_nested_train, y_nested_train, features, className, states, smoothing);
-                // Test model
-                score += clf->score(X_nested_test, y_nested_test);
-            }
-            delete nested_fold;
-            score /= config.nested;
-            if (score > best_fold_score) {
-                best_fold_score = score;
-                best_idx_combination = idx_combination;
-                best_fold_hyper = hyperparam_line;
-            }
-        }
-        delete fold;
-        // Build Classifier with the best hyperparameters to obtain the best score
-        auto hyperparameters = platform::HyperParameters(datasets.getNames(), best_fold_hyper);
-        auto clf = Models::instance()->create(config.model);
-        auto valid = clf->getValidHyperparameters();
-        hyperparameters.check(valid, dataset_name);
-        clf->setHyperparameters(best_fold_hyper);
-        clf->fit(X_train, y_train, features, className, states, smoothing);
-        best_fold_score = clf->score(X_test, y_test);
-        // Return the result
-        result->idx_dataset = task["idx_dataset"].get<int>();
-        result->idx_combination = best_idx_combination;
-        result->score = best_fold_score;
-        result->n_fold = n_fold;
-        result->time = timer.getDuration();
-        // Update progress bar
-        std::cout << get_color_rank(config_mpi.rank) << "*" << std::flush;
-    }
-    json store_result(std::vector<std::string>& names, Task_Result& result, json& results)
-    {
-        json json_result = {
-            { "score", result.score },
-            { "combination", result.idx_combination },
-            { "fold", result.n_fold },
-            { "time", result.time },
-            { "dataset", result.idx_dataset }
-        };
-        auto name = names[result.idx_dataset];
-        if (!results.contains(name)) {
-            results[name] = json::array();
-        }
-        results[name].push_back(json_result);
-        return results;
-    }
-    json producer(std::vector<std::string>& names, json& tasks, struct ConfigMPI& config_mpi, MPI_Datatype& MPI_Result)
-    {
-        Task_Result result;
-        json results;
-        int num_tasks = tasks.size();
-
-        //
-        // 2a.1 Producer will loop to send all the tasks to the consumers and receive the results
-        //
-        for (int i = 0; i < num_tasks; ++i) {
-            MPI_Status status;
-            MPI_Recv(&result, 1, MPI_Result, MPI_ANY_SOURCE, MPI_ANY_TAG, MPI_COMM_WORLD, &status);
-            if (status.MPI_TAG == TAG_RESULT) {
-                //Store result
-                store_result(names, result, results);
-            }
-            MPI_Send(&i, 1, MPI_INT, status.MPI_SOURCE, TAG_TASK, MPI_COMM_WORLD);
-        }
-        //
-        // 2a.2 Producer will send the end message to all the consumers
-        //
-        for (int i = 0; i < config_mpi.n_procs - 1; ++i) {
-            MPI_Status status;
-            MPI_Recv(&result, 1, MPI_Result, MPI_ANY_SOURCE, MPI_ANY_TAG, MPI_COMM_WORLD, &status);
-            if (status.MPI_TAG == TAG_RESULT) {
-                //Store result
-                store_result(names, result, results);
-            }
-            MPI_Send(&i, 1, MPI_INT, status.MPI_SOURCE, TAG_END, MPI_COMM_WORLD);
-        }
-        return results;
-    }
-    void select_best_results_folds(json& results, json& all_results, std::string& model)
-    {
-        Timer timer;
-        auto grid = GridData(Paths::grid_input(model));
-        //
-        // Select the best result of the computed outer folds
-        //
-        for (const auto& result : all_results.items()) {
-            // each result has the results of all the outer folds as each one were a different task
-            double best_score = 0.0;
-            json best;
-            for (const auto& result_fold : result.value()) {
-                double score = result_fold["score"].get<double>();
-                if (score > best_score) {
-                    best_score = score;
-                    best = result_fold;
-                }
-            }
-            auto dataset = result.key();
-            auto combinations = grid.getGrid(dataset);
-            json json_best = {
-                    { "score", best_score },
-                    { "hyperparameters", combinations[best["combination"].get<int>()] },
-                    { "date", get_date() + " " + get_time() },
-                    { "grid", grid.getInputGrid(dataset) },
-                    { "duration", timer.translate2String(best["time"].get<double>()) }
-            };
-            results[dataset] = json_best;
-        }
-    }
-    void consumer(Datasets& datasets, json& tasks, struct ConfigGrid& config, struct ConfigMPI& config_mpi, MPI_Datatype& MPI_Result)
-    {
-        Task_Result result;
-        //
-        // 2b.1 Consumers announce to the producer that they are ready to receive a task
-        //
-        MPI_Send(&result, 1, MPI_Result, config_mpi.manager, TAG_QUERY, MPI_COMM_WORLD);
-        int task;
-        while (true) {
-            MPI_Status status;
-            //
-            // 2b.2 Consumers receive the task from the producer and process it
-            //
-            MPI_Recv(&task, 1, MPI_INT, config_mpi.manager, MPI_ANY_TAG, MPI_COMM_WORLD, &status);
-            if (status.MPI_TAG == TAG_END) {
-                break;
-            }
-            process_task_mpi_consumer(config, config_mpi, tasks, task, datasets, &result);
-            //
-            // 2b.3 Consumers send the result to the producer
-            //
-            MPI_Send(&result, 1, MPI_Result, config_mpi.manager, TAG_RESULT, MPI_COMM_WORLD);
-        }
-    }
-    void GridSearch::go(struct ConfigMPI& config_mpi)
-    {
-        /*
-        * Each task is a json object with the following structure:
-        * {
-        *   "dataset": "dataset_name",
-        *   "idx_dataset": idx_dataset, // used to identify the dataset in the results
-        *    // this index is relative to the used datasets in the actual run not to the whole datasets
-        *   "seed": # of seed to use,
-        *   "Fold": # of fold to process
-        * }
-        *
-        * The overall process consists in these steps:
-           * 0. Create the MPI result type & tasks
-           * 0.1 Create the MPI result type
-           * 0.2 Manager creates the tasks
-           * 1. Manager will broadcast the tasks to all the processes
-           * 1.1 Broadcast the number of tasks
-           * 1.2 Broadcast the length of the following string
-           * 1.2 Broadcast the tasks as a char* string
-           * 2a. Producer delivers the tasks to the consumers
-           * 2a.1 Producer will loop to send all the tasks to the consumers and receive the results
-           * 2a.2 Producer will send the end message to all the consumers
-           * 2b. Consumers process the tasks and send the results to the producer
-           * 2b.1 Consumers announce to the producer that they are ready to receive a task
-           * 2b.2 Consumers receive the task from the producer and process it
-           * 2b.3 Consumers send the result to the producer
-           * 3. Manager select the bests sccores for each dataset
-           * 3.1 Loop thru all the results obtained from each outer fold (task) and select the best
-           * 3.2 Save the results
-        */
-        //
-        // 0.1 Create the MPI result type
-        //
-        Task_Result result;
-        int tasks_size;
-        MPI_Datatype MPI_Result;
-        MPI_Datatype type[5] = { MPI_UNSIGNED, MPI_UNSIGNED, MPI_INT, MPI_DOUBLE, MPI_DOUBLE };
-        int blocklen[5] = { 1, 1, 1, 1, 1 };
-        MPI_Aint disp[5];
-        disp[0] = offsetof(Task_Result, idx_dataset);
-        disp[1] = offsetof(Task_Result, idx_combination);
-        disp[2] = offsetof(Task_Result, n_fold);
-        disp[3] = offsetof(Task_Result, score);
-        disp[4] = offsetof(Task_Result, time);
-        MPI_Type_create_struct(5, blocklen, disp, type, &MPI_Result);
-        MPI_Type_commit(&MPI_Result);
-        //
-        // 0.2 Manager creates the tasks
-        //
-        char* msg;
-        json tasks;
-        if (config_mpi.rank == config_mpi.manager) {
-            timer.start();
-            tasks = build_tasks_mpi(config_mpi.rank);
-            auto tasks_str = tasks.dump();
-            tasks_size = tasks_str.size();
-            msg = new char[tasks_size + 1];
-            strcpy(msg, tasks_str.c_str());
-        }
-        //
-        // 1. Manager will broadcast the tasks to all the processes
-        //
-        MPI_Bcast(&tasks_size, 1, MPI_INT, config_mpi.manager, MPI_COMM_WORLD);
-        if (config_mpi.rank != config_mpi.manager) {
-            msg = new char[tasks_size + 1];
-        }
-        MPI_Bcast(msg, tasks_size + 1, MPI_CHAR, config_mpi.manager, MPI_COMM_WORLD);
-        tasks = json::parse(msg);
-        delete[] msg;
-        auto env = platform::DotEnv();
-        auto datasets = Datasets(config.discretize, Paths::datasets(), env.get("discretize_algo"));
-
-        if (config_mpi.rank == config_mpi.manager) {
-            //
-            // 2a. Producer delivers the tasks to the consumers
-            //
-            auto datasets_names = filterDatasets(datasets);
-            json all_results = producer(datasets_names, tasks, config_mpi, MPI_Result);
-            std::cout << get_color_rank(config_mpi.rank) << separator << std::endl;
-            //
-            // 3. Manager select the bests sccores for each dataset
-            //
-            auto results = initializeResults();
-            select_best_results_folds(results, all_results, config.model);
-            //
-            // 3.2 Save the results
-            //
-            save(results);
-        } else {
-            //
-            // 2b. Consumers process the tasks and send the results to the producer
-            //
-            consumer(datasets, tasks, config, config_mpi, MPI_Result);
-        }
-    }
     json GridSearch::initializeResults()
     {
         // Load previous results if continue is set
         json results;
         if (config.continue_from != NO_CONTINUE()) {
             if (!config.quiet)
-                std::cout << "* Loading previous results" << std::endl;
+                std::cout << Colors::RESET() << "* Loading previous results" << std::endl;
             try {
                 std::ifstream file(Paths::grid_output(config.model));
                 if (file.is_open()) {
@@ -420,4 +93,167 @@ namespace platform {
         };
         file << output.dump(4);
     }
-} /* namespace platform */
+    void GridSearch::compile_results(json& results, json& all_results, std::string& model)
+    {
+        Timer timer;
+        auto grid = GridData(Paths::grid_input(model));
+        //
+        // Select the best result of the computed outer folds
+        //
+        for (const auto& result : all_results.items()) {
+            // each result has the results of all the outer folds as each one were a different task
+            double best_score = 0.0;
+            json best;
+            for (const auto& result_fold : result.value()) {
+                double score = result_fold["score"].get<double>();
+                if (score > best_score) {
+                    best_score = score;
+                    best = result_fold;
+                }
+            }
+            auto dataset = result.key();
+            auto combinations = grid.getGrid(dataset);
+            json json_best = {
+                    { "score", best_score },
+                    { "hyperparameters", combinations[best["combination"].get<int>()] },
+                    { "date", get_date() + " " + get_time() },
+                    { "grid", grid.getInputGrid(dataset) },
+                    { "duration", timer.translate2String(best["time"].get<double>()) }
+            };
+            results[dataset] = json_best;
+        }
+    }
+    json GridSearch::store_result(std::vector<std::string>& names, Task_Result& result, json& results)
+    {
+        json json_result = {
+            { "score", result.score },
+            { "combination", result.idx_combination },
+            { "fold", result.n_fold },
+            { "time", result.time },
+            { "dataset", result.idx_dataset },
+            { "process", result.process },
+            { "task", result.task }
+        };
+        auto name = names[result.idx_dataset];
+        if (!results.contains(name)) {
+            results[name] = json::array();
+        }
+        results[name].push_back(json_result);
+        return results;
+    }
+    void GridSearch::consumer_go(struct ConfigGrid& config, struct ConfigMPI& config_mpi, json& tasks, int n_task, Datasets& datasets, Task_Result* result)
+    {
+        //
+        // initialize
+        //
+        Timer timer;
+        timer.start();
+        json task = tasks[n_task];
+        auto model = config.model;
+        auto grid = GridData(Paths::grid_input(model));
+        auto dataset_name = task["dataset"].get<std::string>();
+        auto idx_dataset = task["idx_dataset"].get<int>();
+        auto seed = task["seed"].get<int>();
+        auto n_fold = task["fold"].get<int>();
+        bool stratified = config.stratified;
+        bayesnet::Smoothing_t smooth;
+        if (config.smooth_strategy == "ORIGINAL")
+            smooth = bayesnet::Smoothing_t::ORIGINAL;
+        else if (config.smooth_strategy == "LAPLACE")
+            smooth = bayesnet::Smoothing_t::LAPLACE;
+        else if (config.smooth_strategy == "CESTNIK")
+            smooth = bayesnet::Smoothing_t::CESTNIK;
+        //
+        // Generate the hyperparameters combinations
+        //
+        auto& dataset = datasets.getDataset(dataset_name);
+        auto combinations = grid.getGrid(dataset_name);
+        dataset.load();
+        auto [X, y] = dataset.getTensors();
+        auto features = dataset.getFeatures();
+        auto className = dataset.getClassName();
+        //
+        // Start working on task
+        //
+        folding::Fold* fold;
+        if (stratified)
+            fold = new folding::StratifiedKFold(config.n_folds, y, seed);
+        else
+            fold = new folding::KFold(config.n_folds, y.size(0), seed);
+        auto [train, test] = fold->getFold(n_fold);
+        auto [X_train, X_test, y_train, y_test] = dataset.getTrainTestTensors(train, test);
+        auto states = dataset.getStates(); // Get the states of the features Once they are discretized
+        float best_fold_score = 0.0;
+        int best_idx_combination = -1;
+        json best_fold_hyper;
+        for (int idx_combination = 0; idx_combination < combinations.size(); ++idx_combination) {
+            auto hyperparam_line = combinations[idx_combination];
+            auto hyperparameters = platform::HyperParameters(datasets.getNames(), hyperparam_line);
+            folding::Fold* nested_fold;
+            if (config.stratified)
+                nested_fold = new folding::StratifiedKFold(config.nested, y_train, seed);
+            else
+                nested_fold = new folding::KFold(config.nested, y_train.size(0), seed);
+            double score = 0.0;
+            for (int n_nested_fold = 0; n_nested_fold < config.nested; n_nested_fold++) {
+                //
+                // Nested level fold
+                //
+                auto [train_nested, test_nested] = nested_fold->getFold(n_nested_fold);
+                auto train_nested_t = torch::tensor(train_nested);
+                auto test_nested_t = torch::tensor(test_nested);
+                auto X_nested_train = X_train.index({ "...", train_nested_t });
+                auto y_nested_train = y_train.index({ train_nested_t });
+                auto X_nested_test = X_train.index({ "...", test_nested_t });
+                auto y_nested_test = y_train.index({ test_nested_t });
+                //
+                // Build Classifier with selected hyperparameters
+                //
+                auto clf = Models::instance()->create(config.model);
+                auto valid = clf->getValidHyperparameters();
+                hyperparameters.check(valid, dataset_name);
+                clf->setHyperparameters(hyperparameters.get(dataset_name));
+                //
+                // Train model
+                //
+                clf->fit(X_nested_train, y_nested_train, features, className, states, smooth);
+                //
+                // Test model
+                //
+                score += clf->score(X_nested_test, y_nested_test);
+            }
+            delete nested_fold;
+            score /= config.nested;
+            if (score > best_fold_score) {
+                best_fold_score = score;
+                best_idx_combination = idx_combination;
+                best_fold_hyper = hyperparam_line;
+            }
+        }
+        delete fold;
+        //
+        // Build Classifier with the best hyperparameters to obtain the best score
+        //
+        auto hyperparameters = platform::HyperParameters(datasets.getNames(), best_fold_hyper);
+        auto clf = Models::instance()->create(config.model);
+        auto valid = clf->getValidHyperparameters();
+        hyperparameters.check(valid, dataset_name);
+        clf->setHyperparameters(best_fold_hyper);
+        clf->fit(X_train, y_train, features, className, states, smooth);
+        best_fold_score = clf->score(X_test, y_test);
+        //
+        // Return the result
+        //
+        result->idx_dataset = task["idx_dataset"].get<int>();
+        result->idx_combination = best_idx_combination;
+        result->score = best_fold_score;
+        result->n_fold = n_fold;
+        result->time = timer.getDuration();
+        result->process = config_mpi.rank;
+        result->task = n_task;
+        //
+        // Update progress bar
+        //
+        std::cout << get_color_rank(config_mpi.rank) << std::flush;
+    }
+} /* namespace platform */

src/grid/GridSearch.h

@@ -4,47 +4,20 @@
 #include <map>
 #include <mpi.h>
 #include <nlohmann/json.hpp>
+#include <folding.hpp>
 #include "common/Datasets.h"
-#include "common/Timer.h"
+#include "common/Timer.hpp"
 #include "main/HyperParameters.h"
 #include "GridData.h"
+#include "GridBase.h"
+#include "bayesnet/network/Network.h"
+
 
 namespace platform {
     using json = nlohmann::ordered_json;
-    struct ConfigGrid {
-        std::string model;
-        std::string score;
-        std::string continue_from;
-        std::string platform;
-        bool quiet;
-        bool only; // used with continue_from to only compute that dataset
-        bool discretize;
-        bool stratified;
-        int nested;
-        int n_folds;
-        json excluded;
-        std::vector<int> seeds;
-    };
-    struct ConfigMPI {
-        int rank;
-        int n_procs;
-        int manager;
-    };
-    typedef struct {
-        uint idx_dataset;
-        uint idx_combination;
-        int n_fold;
-        double score;
-        double time;
-    } Task_Result;
-    const int TAG_QUERY = 1;
-    const int TAG_RESULT = 2;
-    const int TAG_TASK = 3;
-    const int TAG_END = 4;
-    class GridSearch {
+    class GridSearch : public GridBase {
     public:
         explicit GridSearch(struct ConfigGrid& config);
-        void go(struct ConfigMPI& config_mpi);
         ~GridSearch() = default;
         json loadResults();
         static inline std::string NO_CONTINUE() { return "NO_CONTINUE"; }
@@ -52,10 +25,9 @@ namespace platform {
         void save(json& results);
         json initializeResults();
         std::vector<std::string> filterDatasets(Datasets& datasets) const;
-        struct ConfigGrid config;
-        json build_tasks_mpi(int rank);
-        Timer timer; // used to measure the time of the whole process
-        const std::string separator = "|";
+        void compile_results(json& results, json& all_results, std::string& model);
+        json store_result(std::vector<std::string>& names, Task_Result& result, json& results);
+        void consumer_go(struct ConfigGrid& config, struct ConfigMPI& config_mpi, json& tasks, int n_task, Datasets& datasets, Task_Result* result);
     };
 } /* namespace platform */
 #endif

src/main/ArgumentsExperiment.cpp

@@ -0,0 +1,262 @@
+#include "common/Datasets.h"
+#include "common/DotEnv.h"
+#include "common/Paths.h"
+#include "main/Models.h"
+#include "main/modelRegister.h"
+#include "ArgumentsExperiment.h"
+namespace platform {
+    ArgumentsExperiment::ArgumentsExperiment(argparse::ArgumentParser& program, experiment_t type) : arguments{ program }, type{ type }
+    {
+    }
+    void ArgumentsExperiment::add_arguments()
+    {
+        auto env = platform::DotEnv();
+        auto datasets = platform::Datasets(false, platform::Paths::datasets());
+        auto& group = arguments.add_mutually_exclusive_group(true);
+
+        group.add_argument("-d", "--dataset")
+            .help("Dataset file name: " + datasets.toString())
+            .default_value("all")
+            .action([](const std::string& value) {
+            auto datasets = platform::Datasets(false, platform::Paths::datasets());
+            static std::vector<std::string> choices_datasets(datasets.getNames());
+            choices_datasets.push_back("all");
+            if (find(choices_datasets.begin(), choices_datasets.end(), value) != choices_datasets.end()) {
+                return value;
+            }
+            throw std::runtime_error("Dataset must be one of: " + datasets.toString());
+                }
+            );
+        group.add_argument("--datasets").nargs(1, 50).help("Datasets file names 1..50 separated by spaces").default_value(std::vector<std::string>());
+        group.add_argument("--datasets-file").default_value("").help("Datasets file name. Mutually exclusive with dataset. This file should contain a list of datasets to test.");
+        arguments.add_argument("--hyperparameters").default_value("{}").help("Hyperparameters passed to the model in Experiment");
+        arguments.add_argument("--hyper-file").default_value("").help("Hyperparameters file name." \
+            "Mutually exclusive with hyperparameters. This file should contain hyperparameters for each dataset in json format.");
+        arguments.add_argument("--hyper-best").default_value(false).help("Use best results of the model as source of hyperparameters").implicit_value(true);
+        arguments.add_argument("-m", "--model")
+            .help("Model to use: " + platform::Models::instance()->toString())
+            .action([](const std::string& value) {
+            static const std::vector<std::string> choices = platform::Models::instance()->getNames();
+            if (find(choices.begin(), choices.end(), value) != choices.end()) {
+                return value;
+            }
+            throw std::runtime_error("Model must be one of " + platform::Models::instance()->toString());
+                }
+            );
+        arguments.add_argument("--title").default_value("").help("Experiment title");
+        arguments.add_argument("--folder").help("Results folder to use").default_value(platform::Paths::results());
+        arguments.add_argument("--discretize").help("Discretize input dataset").default_value((bool)stoi(env.get("discretize"))).implicit_value(true);
+        auto valid_choices = env.valid_tokens("discretize_algo");
+        auto& disc_arg = arguments.add_argument("--discretize-algo").help("Algorithm to use in discretization. Valid values: " + env.valid_values("discretize_algo")).default_value(env.get("discretize_algo"));
+        for (auto choice : valid_choices) {
+            disc_arg.choices(choice);
+        }
+        valid_choices = env.valid_tokens("smooth_strat");
+        auto& smooth_arg = arguments.add_argument("--smooth-strat").help("Smooth strategy used in Bayes Network node initialization. Valid values: " + env.valid_values("smooth_strat")).default_value(env.get("smooth_strat"));
+        for (auto choice : valid_choices) {
+            smooth_arg.choices(choice);
+        }
+        auto& score_arg = arguments.add_argument("-s", "--score").help("Score to use. Valid values: " + env.valid_values("score")).default_value(env.get("score"));
+        valid_choices = env.valid_tokens("score");
+        for (auto choice : valid_choices) {
+            score_arg.choices(choice);
+        }
+        arguments.add_argument("--no-train-score").help("Don't compute train score").default_value(false).implicit_value(true);
+        arguments.add_argument("--quiet").help("Don't display detailed progress").default_value(false).implicit_value(true);
+        arguments.add_argument("--save").help("Save result (always save even if a dataset is supplied)").default_value(false).implicit_value(true);
+        arguments.add_argument("--stratified").help("If Stratified KFold is to be done").default_value((bool)stoi(env.get("stratified"))).implicit_value(true);
+        arguments.add_argument("-f", "--folds").help("Number of folds").default_value(stoi(env.get("n_folds"))).scan<'i', int>().action([](const std::string& value) {
+            try {
+                auto k = stoi(value);
+                if (k < 2) {
+                    throw std::runtime_error("Number of folds must be greater than 1");
+                }
+                return k;
+            }
+            catch (const runtime_error& err) {
+                throw std::runtime_error(err.what());
+            }
+            catch (...) {
+                throw std::runtime_error("Number of folds must be an integer");
+            }});
+            auto seed_values = env.getSeeds();
+            arguments.add_argument("--seeds").nargs(1, 10).help("Random seeds. Set to -1 to have pseudo random").scan<'i', int>().default_value(seed_values);
+            if (type == experiment_t::NORMAL) {
+                arguments.add_argument("--generate-fold-files").help("generate fold information in datasets_experiment folder").default_value(false).implicit_value(true);
+                arguments.add_argument("--graph").help("generate graphviz dot files with the model").default_value(false).implicit_value(true);
+            }
+    }
+    void ArgumentsExperiment::parse_args(int argc, char** argv)
+    {
+        try {
+            arguments.parse_args(argc, argv);
+        }
+        catch (const exception& err) {
+            cerr << err.what() << std::endl;
+            cerr << arguments;
+            exit(1);
+        }
+        parse();
+    }
+
+    void ArgumentsExperiment::parse()
+    {
+        try {
+            file_name = arguments.get<std::string>("dataset");
+            file_names = arguments.get<std::vector<std::string>>("datasets");
+            datasets_file = arguments.get<std::string>("datasets-file");
+            path_results = arguments.get<std::string>("folder");
+            if (path_results.back() != '/') {
+                path_results += '/';
+            }
+            model_name = arguments.get<std::string>("model");
+            discretize_dataset = arguments.get<bool>("discretize");
+            discretize_algo = arguments.get<std::string>("discretize-algo");
+            smooth_strat = arguments.get<std::string>("smooth-strat");
+            stratified = arguments.get<bool>("stratified");
+            quiet = arguments.get<bool>("quiet");
+            n_folds = arguments.get<int>("folds");
+            score = arguments.get<std::string>("score");
+            seeds = arguments.get<std::vector<int>>("seeds");
+            auto hyperparameters = arguments.get<std::string>("hyperparameters");
+            hyperparameters_json = json::parse(hyperparameters);
+            hyperparameters_file = arguments.get<std::string>("hyper-file");
+            no_train_score = arguments.get<bool>("no-train-score");
+            hyper_best = arguments.get<bool>("hyper-best");
+            if (hyper_best) {
+                // Build the best results file_name
+                hyperparameters_file = path_results + platform::Paths::bestResultsFile(score, model_name);
+                // ignore this parameter
+                hyperparameters = "{}";
+            } else {
+                if (hyperparameters_file != "" && hyperparameters != "{}") {
+                    throw runtime_error("hyperparameters and hyper_file are mutually exclusive");
+                }
+            }
+            title = arguments.get<std::string>("title");
+            if (title == "" && file_name == "all") {
+                throw runtime_error("title is mandatory if all datasets are to be tested");
+            }
+            saveResults = arguments.get<bool>("save");
+            if (type == experiment_t::NORMAL) {
+                graph = arguments.get<bool>("graph");
+                generate_fold_files = arguments.get<bool>("generate-fold-files");
+            } else {
+                graph = false;
+                generate_fold_files = false;
+            }
+        }
+        catch (const exception& err) {
+            cerr << err.what() << std::endl;
+            cerr << arguments;
+            exit(1);
+        }
+        auto datasets = platform::Datasets(false, platform::Paths::datasets());
+        if (datasets_file != "") {
+            ifstream catalog(datasets_file);
+            if (catalog.is_open()) {
+                std::string line;
+                while (getline(catalog, line)) {
+                    if (line.empty() || line[0] == '#') {
+                        continue;
+                    }
+                    if (!datasets.isDataset(line)) {
+                        cerr << "Dataset " << line << " not found" << std::endl;
+                        exit(1);
+                    }
+                    filesToTest.push_back(line);
+                }
+                catalog.close();
+                saveResults = true;
+                if (title == "") {
+                    title = "Test " + to_string(filesToTest.size()) + " datasets (" + datasets_file + ") "\
+                        + model_name + " " + to_string(n_folds) + " folds";
+                }
+            } else {
+                throw std::invalid_argument("Unable to open catalog file. [" + datasets_file + "]");
+            }
+        } else {
+            if (file_names.size() > 0) {
+                for (auto file : file_names) {
+                    if (!datasets.isDataset(file)) {
+                        cerr << "Dataset " << file << " not found" << std::endl;
+                        exit(1);
+                    }
+                }
+                filesToTest = file_names;
+                sort(filesToTest.begin(), filesToTest.end(), [](const auto& lhs, const auto& rhs) {
+                    const auto result = mismatch(lhs.cbegin(), lhs.cend(), rhs.cbegin(), rhs.cend(), [](const auto& lhs, const auto& rhs) {return tolower(lhs) == tolower(rhs);});
+
+                    return result.second != rhs.cend() && (result.first == lhs.cend() || tolower(*result.first) < tolower(*result.second));
+                    });
+                saveResults = true;
+                if (title == "") {
+                    title = "Test " + to_string(file_names.size()) + " datasets " + model_name + " " + to_string(n_folds) + " folds";
+                }
+            } else {
+                if (file_name != "all") {
+                    if (!datasets.isDataset(file_name)) {
+                        cerr << "Dataset " << file_name << " not found" << std::endl;
+                        exit(1);
+                    }
+                    if (title == "") {
+                        title = "Test " + file_name + " " + model_name + " " + to_string(n_folds) + " folds";
+                    }
+                    filesToTest.push_back(file_name);
+                } else {
+                    filesToTest = datasets.getNames();
+                    saveResults = true;
+                }
+            }
+        }
+        if (hyperparameters_file != "") {
+            test_hyperparams = platform::HyperParameters(datasets.getNames(), hyperparameters_file, hyper_best);
+        } else {
+            test_hyperparams = platform::HyperParameters(datasets.getNames(), hyperparameters_json);
+        }
+    }
+    std::string getGppVersion()
+    {
+        std::string result;
+        std::array<char, 128> buffer;
+
+        // Run g++ --version and capture the output
+        using pclose_t = int(*)(FILE*);
+        std::unique_ptr<FILE, pclose_t> pipe(popen("g++ --version", "r"), pclose);
+
+        if (!pipe) {
+            return "Error executing g++ --version command";
+        }
+
+        // Read the first line of output (which contains the version info)
+        if (fgets(buffer.data(), buffer.size(), pipe.get()) != nullptr) {
+            result = buffer.data();
+            // Remove trailing newline if present
+            if (!result.empty() && result[result.length() - 1] == '\n') {
+                result.erase(result.length() - 1);
+            }
+        } else {
+            return "No output from g++ --version command";
+        }
+
+        return result;
+    }
+    Experiment& ArgumentsExperiment::initializedExperiment()
+    {
+        auto env = platform::DotEnv();
+        experiment.setTitle(title).setLanguage("c++").setLanguageVersion(getGppVersion());
+        experiment.setDiscretizationAlgorithm(discretize_algo).setSmoothSrategy(smooth_strat);
+        experiment.setDiscretized(discretize_dataset).setModel(model_name).setPlatform(env.get("platform"));
+        experiment.setStratified(stratified).setNFolds(n_folds).setScoreName(score);
+        experiment.setHyperparameters(test_hyperparams);
+        for (auto seed : seeds) {
+            experiment.addRandomSeed(seed);
+        }
+        experiment.setFilesToTest(filesToTest);
+        experiment.setQuiet(quiet);
+        experiment.setNoTrainScore(no_train_score);
+        experiment.setGenerateFoldFiles(generate_fold_files);
+        experiment.setGraph(graph);
+        return experiment;
+    }
+}

src/main/ArgumentsExperiment.h

@@ -0,0 +1,41 @@
+#ifndef ARGUMENTSEXPERIMENT_H
+#define ARGUMENTSEXPERIMENT_H
+#include <string>
+#include <iostream>
+#include <vector>
+#include <argparse/argparse.hpp>
+#include <nlohmann/json.hpp>
+#include "Experiment.h"
+
+namespace platform {
+    using json = nlohmann::ordered_json;
+    enum class experiment_t { NORMAL, GRID };
+    class ArgumentsExperiment {
+    public:
+        ArgumentsExperiment(argparse::ArgumentParser& program, experiment_t type);
+        ~ArgumentsExperiment() = default;
+        std::vector<std::string> getFilesToTest() const { return filesToTest; }
+        void add_arguments();
+        void parse_args(int argc, char** argv);
+        void parse();
+        Experiment& initializedExperiment();
+        bool isQuiet() const { return quiet; }
+        bool haveToSaveResults() const { return saveResults; }
+        bool doGraph() const { return graph; }
+        std::string getPathResults() const { return path_results; }
+    private:
+        Experiment experiment;
+        experiment_t type;
+        argparse::ArgumentParser& arguments;
+        std::string file_name, model_name, title, hyperparameters_file, datasets_file, discretize_algo, smooth_strat;
+        std::string score, path_results;
+        json hyperparameters_json;
+        bool discretize_dataset, stratified, saveResults, quiet, no_train_score, generate_fold_files, graph, hyper_best;
+        std::vector<int> seeds;
+        std::vector<std::string> file_names;
+        std::vector<std::string> filesToTest;
+        platform::HyperParameters test_hyperparams;
+        int n_folds;
+    };
+}
+#endif

src/main/Experiment.cpp

@@ -7,16 +7,18 @@
 namespace platform {
     using json = nlohmann::ordered_json;
 
-    void Experiment::saveResult()
+    void Experiment::saveResult(const std::string& path)
     {
-        result.save();
-        std::cout << "Result saved in " << Paths::results() << result.getFilename() << std::endl;
+        result.setSchemaVersion("1.0");
+        result.check();
+        result.save(path);
+        std::cout << "Result saved in " << path << result.getFilename() << std::endl;
     }
-    void Experiment::report(bool classification_report)
+    void Experiment::report()
     {
         ReportConsole report(result.getJson());
         report.show();
-        if (classification_report) {
+        if (filesToTest.size() == 1) {
             std::cout << report.showClassificationReport(Colors::BLUE());
         }
     }
@@ -41,9 +43,25 @@ namespace platform {
             }
         }
     }
-    void Experiment::go(std::vector<std::string> filesToProcess, bool quiet, bool no_train_score, bool generate_fold_files, bool graph)
+    Experiment& Experiment::setSmoothSrategy(const std::string& smooth_strategy)
     {
-        for (auto fileName : filesToProcess) {
+        this->smooth_strategy = smooth_strategy;
+        this->result.setSmoothStrategy(smooth_strategy);
+        if (smooth_strategy == "ORIGINAL")
+            smooth_type = bayesnet::Smoothing_t::ORIGINAL;
+        else if (smooth_strategy == "LAPLACE")
+            smooth_type = bayesnet::Smoothing_t::LAPLACE;
+        else if (smooth_strategy == "CESTNIK")
+            smooth_type = bayesnet::Smoothing_t::CESTNIK;
+        else {
+            std::cerr << "Experiment: Unknown smoothing strategy: " << smooth_strategy << std::endl;
+            exit(1);
+        }
+        return *this;
+    }
+    void Experiment::go()
+    {
+        for (auto fileName : filesToTest) {
             if (fileName.size() > max_name)
                 max_name = fileName.size();
         }
@@ -58,14 +76,16 @@ namespace platform {
             std::cout << " ( " << Colors::GREEN() << "b" << Colors::RESET() << " )  Scoring train dataset" << std::endl;
             std::cout << " ( " << Colors::GREEN() << "c" << Colors::RESET() << " )  Scoring test dataset" << std::endl << std::endl;
             std::cout << Colors::YELLOW() << "Note: fold number in this color means fitting had issues such as not using all features in BoostAODE classifier" << std::endl << std::endl;
-            std::cout << Colors::GREEN() << left << "  #  " << setw(max_name) << "Dataset" << " #Samp #Feat Seed Status" << string(3 * nfolds - 2, ' ') << " Time" << std::endl;
-            std::cout << " --- " << string(max_name, '-') << " ----- ----- ---- " << string(4 + 3 * nfolds, '-') << " ----------" << Colors::RESET() << std::endl;
+            int nc = 4 + 3 * nfolds + (nfolds >= 10 ? nfolds - 10 + 1 : 0);
+            std::cout << Colors::GREEN() << left << "  #  " << setw(max_name) << "Dataset" << " #Samp #Feat Seed Status" << string(nc - 6, ' ') << setw(11) << " Time" << " Score" << std::endl;
+            std::cout << " --- " << string(max_name, '-') << " ----- ----- ---- " << string(nc, '-') << " ----------" << " ---------";
+            std::cout << Colors::RESET() << std::endl;
         }
         int num = 0;
-        for (auto fileName : filesToProcess) {
+        for (auto fileName : filesToTest) {
             if (!quiet)
                 std::cout << " " << setw(3) << right << num++ << " " << setw(max_name) << left << fileName << right << flush;
-            cross_validation(fileName, quiet, no_train_score, generate_fold_files, graph);
+            cross_validation(fileName);
             if (!quiet)
                 std::cout << std::endl;
         }
@@ -95,7 +115,8 @@ namespace platform {
     }
     void showProgress(int fold, const std::string& color, const std::string& phase)
     {
-        std::string prefix = phase == "-" ? "" : "\b\b\b\b";
+        int nc = fold >= 10 ? 5 : 4;
+        std::string prefix = phase == "-" ? "" : std::string(nc, '\b');
         std::cout << prefix << color << fold << Colors::RESET() << "(" << color << phase << Colors::RESET() << ")" << flush;
 
     }
@@ -137,7 +158,7 @@ namespace platform {
         file << output.dump(4);
         file.close();
     }
-    void Experiment::cross_validation(const std::string& fileName, bool quiet, bool no_train_score, bool generate_fold_files, bool graph)
+    void Experiment::cross_validation(const std::string& fileName)
     {
         //
         // Load dataset and prepare data
@@ -224,11 +245,9 @@ namespace platform {
                 // Train model
                 //
                 clf->fit(X_train, y_train, features, className, states, smooth_type);
-                if (!quiet)
-                    showProgress(nfold + 1, getColor(clf->getStatus()), "b");
                 auto clf_notes = clf->getNotes();
-                std::transform(clf_notes.begin(), clf_notes.end(), std::back_inserter(notes), [nfold](const std::string& note)
-                    { return "Fold " + std::to_string(nfold) + ": " + note; });
+                std::transform(clf_notes.begin(), clf_notes.end(), std::back_inserter(notes), [seed, nfold](const std::string& note)
+                    { return "Seed: " + std::to_string(seed) + " Fold: " + std::to_string(nfold) + ": " + note; });
                 nodes[item] = clf->getNumberOfNodes();
                 edges[item] = clf->getNumberOfEdges();
                 num_states[item] = clf->getNumberOfStates();
@@ -238,10 +257,13 @@ namespace platform {
                 // Score train
                 //
                 if (!no_train_score) {
+                    if (!quiet)
+                        showProgress(nfold + 1, getColor(clf->getStatus()), "b");
                     auto y_proba_train = clf->predict_proba(X_train);
                     Scores scores(y_train, y_proba_train, num_classes, labels);
                     score_train_value = score == score_t::ACCURACY ? scores.accuracy() : scores.auc();
-                    confusion_matrices_train.push_back(scores.get_confusion_matrix_json(true));
+                    if (discretized)
+                        confusion_matrices_train.push_back(scores.get_confusion_matrix_json(true));
                 }
                 //
                 // Test model
@@ -256,7 +278,8 @@ namespace platform {
                 test_time[item] = test_timer.getDuration();
                 score_train[item] = score_train_value;
                 score_test[item] = score_test_value;
-                confusion_matrices.push_back(scores.get_confusion_matrix_json(true));
+                if (discretized)
+                    confusion_matrices.push_back(scores.get_confusion_matrix_json(true));
                 if (!quiet)
                     std::cout << "\b\b\b, " << flush;
                 //
@@ -277,10 +300,13 @@ namespace platform {
             }
             if (!quiet) {
                 seed_timer.stop();
-                std::cout << "end. [" << seed_timer.getDurationString() << "]" << std::endl;
+                std::cout << "end. " << std::setw(10) << std::right << seed_timer.getDurationString();
             }
             delete fold;
         }
+        // Show Results
+        if (!quiet)
+            std::cout << " " << setw(9) << right << std::fixed << std::setprecision(7) << torch::mean(score_test).item<double>();
         //
         // Store result totals in Result
         //

src/main/Experiment.h

@@ -25,21 +25,7 @@ namespace platform {
         {
             this->discretization_algo = discretization_algo; this->result.setDiscretizationAlgorithm(discretization_algo); return *this;
         }
-        Experiment& setSmoothSrategy(const std::string& smooth_strategy)
-        {
-            this->smooth_strategy = smooth_strategy; this->result.setSmoothStrategy(smooth_strategy);
-            if (smooth_strategy == "ORIGINAL")
-                smooth_type = bayesnet::Smoothing_t::ORIGINAL;
-            else if (smooth_strategy == "LAPLACE")
-                smooth_type = bayesnet::Smoothing_t::LAPLACE;
-            else if (smooth_strategy == "CESTNIK")
-                smooth_type = bayesnet::Smoothing_t::CESTNIK;
-            else {
-                std::cerr << "Experiment: Unknown smoothing strategy: " << smooth_strategy << std::endl;
-                exit(1);
-            }
-            return *this;
-        }
+        Experiment& setSmoothSrategy(const std::string& smooth_strategy);
         Experiment& setLanguageVersion(const std::string& language_version) { this->result.setLanguageVersion(language_version); return *this; }
         Experiment& setDiscretized(bool discretized) { this->discretized = discretized; result.setDiscretized(discretized); return *this; }
         Experiment& setStratified(bool stratified) { this->stratified = stratified; result.setStratified(stratified); return *this; }
@@ -48,18 +34,33 @@ namespace platform {
         Experiment& addRandomSeed(int randomSeed) { randomSeeds.push_back(randomSeed); result.addSeed(randomSeed); return *this; }
         Experiment& setDuration(float duration) { this->result.setDuration(duration); return *this; }
         Experiment& setHyperparameters(const HyperParameters& hyperparameters_) { this->hyperparameters = hyperparameters_; return *this; }
-        void cross_validation(const std::string& fileName, bool quiet, bool no_train_score, bool generate_fold_files, bool graph);
-        void go(std::vector<std::string> filesToProcess, bool quiet, bool no_train_score, bool generate_fold_files, bool graph);
-        void saveResult();
+        HyperParameters& getHyperParameters() { return hyperparameters; }
+        std::string getModel() const { return result.getModel(); }
+        std::string getScore() const { return result.getScoreName(); }
+        bool isDiscretized() const { return discretized; }
+        bool isStratified() const { return stratified; }
+        bool isQuiet() const { return quiet; }
+        std::string getSmoothStrategy() const { return smooth_strategy; }
+        int getNFolds() const { return nfolds; }
+        std::vector<int> getRandomSeeds() const { return randomSeeds; }
+        void cross_validation(const std::string& fileName);
+        void go();
+        void saveResult(const std::string& path);
         void show();
         void saveGraph();
-        void report(bool classification_report = false);
+        void report();
+        void setFilesToTest(const std::vector<std::string>& filesToTest) { this->filesToTest = filesToTest; }
+        void setQuiet(bool quiet) { this->quiet = quiet; }
+        void setNoTrainScore(bool no_train_score) { this->no_train_score = no_train_score; }
+        void setGenerateFoldFiles(bool generate_fold_files) { this->generate_fold_files = generate_fold_files; }
+        void setGraph(bool graph) { this->graph = graph; }
     private:
         score_t parse_score() const;
         Result result;
-        bool discretized{ false }, stratified{ false };
+        bool discretized{ false }, stratified{ false }, generate_fold_files{ false }, graph{ false }, quiet{ false }, no_train_score{ false };
         std::vector<PartialResult> results;
         std::vector<int> randomSeeds;
+        std::vector<std::string> filesToTest;
         std::string discretization_algo;
         std::string smooth_strategy;
         bayesnet::Smoothing_t smooth_type{ bayesnet::Smoothing_t::NONE };

src/main/Models.h

@@ -5,11 +5,15 @@
 #include <bayesnet/ensembles/AODE.h>
 #include <bayesnet/ensembles/A2DE.h>
 #include <bayesnet/ensembles/AODELd.h>
+#include <bayesnet/ensembles/XBAODE.h>
+#include <bayesnet/ensembles/XBA2DE.h>
 #include <bayesnet/ensembles/BoostAODE.h>
 #include <bayesnet/ensembles/BoostA2DE.h>
 #include <bayesnet/classifiers/TAN.h>
 #include <bayesnet/classifiers/KDB.h>
 #include <bayesnet/classifiers/SPODE.h>
+#include <bayesnet/classifiers/XSPODE.h>
+#include <bayesnet/classifiers/XSP2DE.h>
 #include <bayesnet/classifiers/SPnDE.h>
 #include <bayesnet/classifiers/TANLd.h>
 #include <bayesnet/classifiers/KDBLd.h>
@@ -19,7 +23,12 @@
 #include <pyclassifiers/ODTE.h>
 #include <pyclassifiers/SVC.h>
 #include <pyclassifiers/XGBoost.h>
+#include <pyclassifiers/AdaBoostPy.h>
 #include <pyclassifiers/RandomForest.h>
+#include "../experimental_clfs/XA1DE.h"
+#include "../experimental_clfs/AdaBoost.h"
+#include "../experimental_clfs/DecisionTree.h"
+
 namespace platform {
     class Models {
     public:
@@ -42,4 +51,4 @@ namespace platform {
         Registrar(const std::string& className, function<bayesnet::BaseClassifier* (void)> classFactoryFunction);
     };
 }
-#endif
+#endif

src/main/RocAuc.cpp

@@ -3,8 +3,9 @@
 #include <numeric>
 #include <utility>
 #include "RocAuc.h"
+#include "common/TensorUtils.h" // tensorToVector
 namespace platform {
-    
+
     double RocAuc::compute(const torch::Tensor& y_proba, const torch::Tensor& labels)
     {
         size_t nClasses = y_proba.size(1);
@@ -48,6 +49,7 @@ namespace platform {
         double tp = 0, fp = 0;
         double totalPos = std::count(y_test.begin(), y_test.end(), classIdx);
         double totalNeg = nSamples - totalPos;
+        if (totalPos == 0 || totalNeg == 0) return 0.5; // neutral AUC
 
         for (const auto& [score, label] : scoresAndLabels) {
             if (label == 1) {

src/main/Scores.cpp

@@ -1,6 +1,6 @@
 #include <sstream>
 #include "Scores.h"
-#include "common/Utils.h" // tensorToVector
+#include "common/TensorUtils.h" // tensorToVector
 #include "common/Colors.h"
 namespace platform {
     Scores::Scores(torch::Tensor& y_test, torch::Tensor& y_proba, int num_classes, std::vector<std::string> labels) : num_classes(num_classes), labels(labels), y_test(y_test), y_proba(y_proba)

src/main/modelRegister.h

@@ -1,39 +1,55 @@
 #ifndef MODELREGISTER_H
 #define MODELREGISTER_H
-
-static platform::Registrar registrarT("TAN",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::TAN();});
-static platform::Registrar registrarTLD("TANLd",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::TANLd();});
-static platform::Registrar registrarS("SPODE",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::SPODE(2);});
-static platform::Registrar registrarSn("SPnDE",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::SPnDE({ 0, 1 });});
-static platform::Registrar registrarSLD("SPODELd",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::SPODELd(2);});
-static platform::Registrar registrarK("KDB",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::KDB(2);});
-static platform::Registrar registrarKLD("KDBLd",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::KDBLd(2);});
-static platform::Registrar registrarA("AODE",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::AODE();});
-static platform::Registrar registrarA2("A2DE",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::A2DE();});
-static platform::Registrar registrarALD("AODELd",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::AODELd();});
-static platform::Registrar registrarBA("BoostAODE",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::BoostAODE();});
-static platform::Registrar registrarBA2("BoostA2DE",
-    [](void) -> bayesnet::BaseClassifier* { return new bayesnet::BoostA2DE();});
-static platform::Registrar registrarSt("STree",
-    [](void) -> bayesnet::BaseClassifier* { return new pywrap::STree();});
-static platform::Registrar registrarOdte("Odte",
-    [](void) -> bayesnet::BaseClassifier* { return new pywrap::ODTE();});
-static platform::Registrar registrarSvc("SVC",
-    [](void) -> bayesnet::BaseClassifier* { return new pywrap::SVC();});
-static platform::Registrar registrarRaF("RandomForest",
-    [](void) -> bayesnet::BaseClassifier* { return new pywrap::RandomForest();});
-static platform::Registrar registrarXGB("XGBoost",
-    [](void) -> bayesnet::BaseClassifier* { return new pywrap::XGBoost();});
-
-#endif
+namespace platform {
+    static Registrar registrarT("TAN",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::TAN();});
+    static Registrar registrarTLD("TANLd",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::TANLd();});
+    static Registrar registrarS("SPODE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::SPODE(2);});
+    static Registrar registrarSn("SPnDE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::SPnDE({ 0, 1 });});
+    static Registrar registrarSLD("SPODELd",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::SPODELd(2);});
+    static Registrar registrarK("KDB",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::KDB(2);});
+    static Registrar registrarKLD("KDBLd",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::KDBLd(2);});
+    static Registrar registrarA("AODE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::AODE();});
+    static Registrar registrarA2("A2DE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::A2DE();});
+    static Registrar registrarALD("AODELd",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::AODELd();});
+    static Registrar registrarBA("BoostAODE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::BoostAODE();});
+    static Registrar registrarBA2("BoostA2DE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::BoostA2DE();});
+    static Registrar registrarSt("STree",
+        [](void) -> bayesnet::BaseClassifier* { return new pywrap::STree();});
+    static Registrar registrarOdte("Odte",
+        [](void) -> bayesnet::BaseClassifier* { return new pywrap::ODTE();});
+    static Registrar registrarSvc("SVC",
+        [](void) -> bayesnet::BaseClassifier* { return new pywrap::SVC();});
+    static Registrar registrarRaF("RandomForest",
+        [](void) -> bayesnet::BaseClassifier* { return new pywrap::RandomForest();});
+    static Registrar registrarXGB("XGBoost",
+        [](void) -> bayesnet::BaseClassifier* { return new pywrap::XGBoost();});
+    static Registrar registrarAdaPy("AdaBoostPy",
+        [](void) -> bayesnet::BaseClassifier* { return new pywrap::AdaBoostPy();});
+    static Registrar registrarAda("AdaBoost",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::AdaBoost();});
+    static Registrar registrarDT("DecisionTree",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::DecisionTree();});
+    static Registrar registrarXSPODE("XSPODE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::XSpode(0);});
+    static Registrar registrarXSP2DE("XSP2DE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::XSp2de(0, 1);});
+    static Registrar registrarXBAODE("XBAODE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::XBAODE();});
+    static Registrar registrarXBA2DE("XBA2DE",
+        [](void) -> bayesnet::BaseClassifier* { return new bayesnet::XBA2DE();});
+    static Registrar registrarXA1DE("XA1DE",
+        [](void) -> bayesnet::BaseClassifier* { return new XA1DE();});
+}
+#endif

src/manage/ManageScreen.cpp

@@ -18,8 +18,8 @@ namespace platform {
     const std::string STATUS_OK = "Ok.";
     const std::string STATUS_COLOR = Colors::GREEN();
 
-    ManageScreen::ManageScreen(int rows, int cols, const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial, bool compare) :
-        rows{ rows }, cols{ cols }, complete{ complete }, partial{ partial }, compare{ compare }, didExcel(false), results(ResultsManager(model, score, platform, complete, partial))
+    ManageScreen::ManageScreen(const std::string path_, int rows, int cols, const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial, bool compare) :
+        path{ path_ }, rows{ rows }, cols{ cols }, complete{ complete }, partial{ partial }, compare{ compare }, didExcel(false), results(ResultsManager(path_, model, score, platform, complete, partial))
     {
         results.load();
         openExcel = false;
@@ -82,10 +82,12 @@ namespace platform {
             workbook_close(workbook);
         }
         if (didExcel) {
-            std::cout << Colors::MAGENTA() << "Excel file created: " << Paths::excel() + Paths::excelResults() << std::endl;
+            excelFileName = Paths::excel() + Paths::excelResults();
+            std::cout << Colors::MAGENTA() << "Excel file created: " << excelFileName << std::endl;
         }
         std::cout << Colors::RESET() << "Done!" << std::endl;
     }
+
     std::string ManageScreen::getVersions()
     {
         std::string kfold_version = folding::KFold(5, 100).version();
@@ -257,8 +259,9 @@ namespace platform {
         auto [index_from, index_to] = paginator[static_cast<int>(output_type)].getOffset();
         for (int i = index_from; i <= index_to; i++) {
             auto color = (i % 2) ? Colors::BLUE() : Colors::CYAN();
-            std::cout << color << std::setw(3) << std::fixed << std::right << i << " ";
-            std::cout << results.at(i).to_string(maxModel, maxTitle) << std::endl;
+            auto color_status = results.at(i).check().size() == 0 ? color : Colors::RED();
+            std::cout << color_status << std::setw(3) << std::fixed << std::right << i << " ";
+            std::cout << color << results.at(i).to_string(maxModel, maxTitle) << std::endl;
         }
         //
         // Status Area
@@ -311,6 +314,34 @@ namespace platform {
             return "Reporting " + results.at(index).getFilename();
         }
     }
+    void ManageScreen::changeModel(const int index)
+    {
+        std::cout << "Old model: " << results.at(index).getModel() << std::endl;
+        std::cout << "New model: ";
+        std::string newModel;
+        getline(std::cin, newModel);
+        if (newModel.empty()) {
+            list("Model not changed", Colors::YELLOW());
+            return;
+        }
+        if (newModel == results.at(index).getModel()) {
+            list("Model already set to " + newModel, Colors::RED());
+            return;
+        }
+        // Remove the old result file
+        std::string oldFile = path + results.at(index).getFilename();
+        std::filesystem::remove(oldFile);
+        // Actually change the model
+        results.at(index).setModel(newModel);
+        results.at(index).save(path);
+        int newModelSize = static_cast<int>(newModel.size());
+        if (newModelSize > maxModel) {
+            maxModel = newModelSize;
+            header_lengths[2] = maxModel;
+            updateSize(rows, cols);
+        }
+        list("Model changed to " + newModel, Colors::GREEN());
+    }
     std::pair<std::string, std::string> ManageScreen::sortList()
     {
         std::vector<std::tuple<std::string, char, bool>>  sortOptions = {
@@ -371,6 +402,7 @@ namespace platform {
             {"list", 'l', false},
             {"Delete", 'D', true},
             {"datasets", 'd', false},
+            {"change model", 'm', true},
             {"hide", 'h', true},
             {"sort", 's', false},
             {"report", 'r', true},
@@ -457,20 +489,19 @@ namespace platform {
                     index_A = index;
                     list("A set to " + std::to_string(index), Colors::GREEN());
                     break;
-                case 'B': // set_b or back to list
-                    if (output_type == OutputType::EXPERIMENTS) {
-                        if (index == index_A) {
-                            list("A and B cannot be the same!", Colors::RED());
-                            break;
-                        }
-                        index_B = index;
-                        list("B set to " + std::to_string(index), Colors::GREEN());
-                    } else {
-                        // back to show the report
-                        output_type = OutputType::RESULT;
-                        paginator[static_cast<int>(OutputType::DETAIL)].setPage(1);
-                        list(STATUS_OK, STATUS_COLOR);
+                case 'B': // set_b
+                    if (index == index_A) {
+                        list("A and B cannot be the same!", Colors::RED());
+                        break;
                     }
+                    index_B = index;
+                    list("B set to " + std::to_string(index), Colors::GREEN());
+                    break;
+                case 'b': // back to list
+                    // back to show the report
+                    output_type = OutputType::RESULT;
+                    paginator[static_cast<int>(OutputType::DETAIL)].setPage(1);
+                    list(STATUS_OK, STATUS_COLOR);
                     break;
                 case 'c':
                     if (index_A == -1 || index_B == -1) {
@@ -497,6 +528,9 @@ namespace platform {
                     paginator[static_cast<int>(OutputType::EXPERIMENTS)].setTotal(results.size());
                     list(filename + " deleted!", Colors::RED());
                     break;
+                case 'm':
+                    changeModel(index);
+                    break;
                 case 'h':
                     {
                         std::string status_message;
@@ -543,16 +577,14 @@ namespace platform {
                     break;
                 case 't':
                     {
-                        std::string status_message;
                         std::cout << "Title: " << results.at(index).getTitle() << std::endl;
                         std::cout << "New title: ";
                         std::string newTitle;
                         getline(std::cin, newTitle);
                         if (!newTitle.empty()) {
                             results.at(index).setTitle(newTitle);
-                            results.at(index).save();
-                            status_message = "Title changed to " + newTitle;
-                            list(status_message, Colors::GREEN());
+                            results.at(index).save(path);
+                            list("Title changed to " + newTitle, Colors::GREEN());
                             break;
                         }
                         list("No title change!", Colors::YELLOW());

src/manage/ManageScreen.h

@@ -15,10 +15,11 @@ namespace platform {
     };
     class ManageScreen {
     public:
-        ManageScreen(int rows, int cols, const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial, bool compare);
+        ManageScreen(const std::string path, int rows, int cols, const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial, bool compare);
         ~ManageScreen() = default;
         void doMenu();
         void updateSize(int rows, int cols);
+        std::string getExcelFileName() const { return excelFileName; }
     private:
         void list(const std::string& status, const std::string& color);
         void list_experiments(const std::string& status, const std::string& color);
@@ -27,6 +28,7 @@ namespace platform {
         void list_datasets(const std::string& status, const std::string& color);
         bool confirmAction(const std::string& intent, const std::string& fileName) const;
         std::string report(const int index, const bool excelReport);
+        void changeModel(const int index);
         std::string report_compared();
         std::pair<std::string, std::string> sortList();
         std::string getVersions();
@@ -57,6 +59,7 @@ namespace platform {
         std::vector<Paginator> paginator;
         ResultsManager results;
         lxw_workbook* workbook;
+        std::string path, excelFileName;
     };
 }
 #endif

src/manage/ResultsManager.cpp

@@ -1,10 +1,9 @@
 #include <algorithm>
-#include "common/Paths.h"
 #include "ResultsManager.h"
 
 namespace platform {
-    ResultsManager::ResultsManager(const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial) :
-        path(Paths::results()), model(model), scoreName(score), platform(platform), complete(complete), partial(partial), maxModel(0), maxTitle(0)
+    ResultsManager::ResultsManager(const std::string& path_, const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial) :
+        path(path_), model(model), scoreName(score), platform(platform), complete(complete), partial(partial), maxModel(0), maxTitle(0)
     {
     }
     void ResultsManager::load()

src/manage/ResultsManager.h

@@ -18,7 +18,7 @@ namespace platform {
     };
     class ResultsManager {
     public:
-        ResultsManager(const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial);
+        ResultsManager(const std::string& path_, const std::string& model, const std::string& score, const std::string& platform, bool complete, bool partial);
         void load(); // Loads the list of results
         void sortResults(SortField field, SortType type); // Sorts the list of results
         void sortDate(SortType type);

src/reports/DatasetsConsole.cpp

@@ -26,9 +26,10 @@ namespace platform {
         auto datasets = platform::Datasets(false, platform::Paths::datasets());
         std::stringstream sheader;
         auto datasets_names = datasets.getNames();
+        std::cout << Colors::GREEN() << "Datasets available in the platform: " << datasets_names.size() << std::endl;
         int maxName = std::max(size_t(7), (*max_element(datasets_names.begin(), datasets_names.end(), [](const std::string& a, const std::string& b) { return a.size() < b.size(); })).size());
         std::vector<std::string> header_labels = { " #", "Dataset", "Sampl.", "Feat.", "#Num.", "Cls", "Balance" };
-        std::vector<int> header_lengths = { 3, maxName, 6, 5, 5, 3, DatasetsConsole::BALANCE_LENGTH };
+        std::vector<int> header_lengths = { 3, maxName, 6, 6, 6, 3, DatasetsConsole::BALANCE_LENGTH };
         sheader << Colors::GREEN();
         for (int i = 0; i < header_labels.size(); i++) {
             sheader << setw(header_lengths[i]) << left << header_labels[i] << " ";
@@ -51,19 +52,23 @@ namespace platform {
             auto& dataset = datasets.getDataset(dataset_name);
             dataset.load();
             auto nSamples = dataset.getNSamples();
-            line << setw(6) << right << nSamples << " ";
+            line << setw(header_lengths[2]) << right << nSamples << " ";
             auto nFeatures = dataset.getFeatures().size();
-            line << setw(5) << right << nFeatures << " ";
+            line << setw(header_lengths[3]) << right << nFeatures << " ";
             auto numericFeatures = dataset.getNumericFeatures();
             auto num = std::count(numericFeatures.begin(), numericFeatures.end(), true);
-            line << setw(5) << right << num << " ";
+            line << setw(header_lengths[4]) << right << num << " ";
             auto nClasses = dataset.getNClasses();
-            line << setw(3) << right << nClasses << " ";
+            line << setw(header_lengths[5]) << right << nClasses << " ";
             std::string sep = "";
             oss.str("");
-            for (auto number : dataset.getClassesCounts()) {
-                oss << sep << std::setprecision(2) << fixed << (float)number / nSamples * 100.0 << "% (" << number << ")";
-                sep = " / ";
+            if (nSamples == 0) {
+                oss << "No samples";
+            } else {
+                for (auto number : dataset.getClassesCounts()) {
+                    oss << sep << std::setprecision(2) << fixed << (float)number / nSamples * 100.0 << "% (" << number << ")";
+                    sep = " / ";
+                }
             }
             split_lines(maxName, line.str(), oss.str());
             // Store data for Excel report

src/reports/DatasetsExcel.cpp

@@ -1,8 +1,9 @@
+#include "common/Paths.h"
 #include "DatasetsExcel.h"
 namespace platform {
     DatasetsExcel::DatasetsExcel()
     {
-        file_name = "datasets.xlsx";
+        file_name = Paths::excelDatasets();
         workbook = workbook_new(getFileName().c_str());
         createFormats();
         setProperties("Datasets");

src/reports/DatasetsExcel.h

@@ -11,6 +11,7 @@ namespace platform {
         DatasetsExcel();
         ~DatasetsExcel();
         void report(json& data);
+        std::string getExcelFileName() { return getFileName(); }
     };
 }
 #endif

src/reports/ReportConsole.cpp

@@ -2,7 +2,7 @@
 #include <locale>
 #include "best/BestScore.h"
 #include "common/CLocale.h"
-#include "common/Timer.h"
+#include "common/Timer.hpp"
 #include "ReportConsole.h"
 #include "main/Scores.h"
 
@@ -49,7 +49,8 @@ namespace platform {
         oss << "Execution took  " << timer.translate2String(data["duration"].get<float>())
             << " on " << data["platform"].get<std::string>() << " Language: " << data["language"].get<std::string>();
         sheader << headerLine(oss.str());
-        sheader << headerLine("Score is " + data["score_name"].get<std::string>());
+        std::string schema_version = data.find("schema_version") != data.end() ? data["schema_version"].get<std::string>() : "-";
+        sheader << headerLine("Score is " + data["score_name"].get<std::string>() + " Schema version: " + schema_version);
         sheader << std::string(MAXL, '*') << std::endl;
         sheader << std::endl;
     }
@@ -83,7 +84,7 @@ namespace platform {
         }
         std::vector<std::string> header_labels = { " #", "Dataset", "Sampl.", "Feat.", "Cls", nodes_label, leaves_label, depth_label, "Score", "Time", "Hyperparameters" };
         sheader << Colors::GREEN();
-        std::vector<int> header_lengths = { 3, maxDataset, 6, 5, 3, 9, 9, 9, 15, 20, maxHyper };
+        std::vector<int> header_lengths = { 3, maxDataset, 6, 6, 3, 13, 13, 13, 15, 20, maxHyper };
         for (int i = 0; i < header_labels.size(); i++) {
             sheader << std::setw(header_lengths[i]) << std::left << header_labels[i] << " ";
         }
@@ -107,11 +108,11 @@ namespace platform {
             line << std::setw(3) << std::right << index++ << " ";
             line << std::setw(maxDataset) << std::left << r["dataset"].get<std::string>() << " ";
             line << std::setw(6) << std::right << r["samples"].get<int>() << " ";
-            line << std::setw(5) << std::right << r["features"].get<int>() << " ";
+            line << std::setw(6) << std::right << r["features"].get<int>() << " ";
             line << std::setw(3) << std::right << r["classes"].get<int>() << " ";
-            line << std::setw(9) << std::setprecision(2) << std::fixed << r["nodes"].get<float>() << " ";
-            line << std::setw(9) << std::setprecision(2) << std::fixed << r["leaves"].get<float>() << " ";
-            line << std::setw(9) << std::setprecision(2) << std::fixed << r["depth"].get<float>() << " ";
+            line << std::setw(13) << std::setprecision(2) << std::fixed << r["nodes"].get<float>() << " ";
+            line << std::setw(13) << std::setprecision(2) << std::fixed << r["leaves"].get<float>() << " ";
+            line << std::setw(13) << std::setprecision(2) << std::fixed << r["depth"].get<float>() << " ";
             line << std::setw(8) << std::right << std::setprecision(6) << std::fixed << r["score"].get<double>() << "±" << std::setw(6) << std::setprecision(4) << std::fixed << r["score_std"].get<double>();
             const std::string status = compareResult(r["dataset"].get<std::string>(), r["score"].get<double>());
             line << status;
@@ -223,7 +224,7 @@ namespace platform {
     std::string ReportConsole::buildClassificationReport(json& result, std::string color)
     {
         std::stringstream oss;
-        if (result.find("confusion_matrices") == result.end())
+        if (result.find("confusion_matrices") == result.end() || result["confusion_matrices"].size() == 0)
             return "";
         bool second_header = false;
         int lines_header = 0;

src/results/JsonValidator.h

@@ -0,0 +1,137 @@
+#ifndef JSONVALIDATOR_H
+#define JSONVALIDATOR_H
+#include <fstream>
+#include <vector>
+#include <regex>
+#include <nlohmann/json.hpp>
+
+namespace platform {
+    using json = nlohmann::ordered_json;
+    class JsonValidator {
+    public:
+        JsonValidator(const json& schema) : schema(schema) {}
+
+        std::vector<std::string> validate_file(const std::string& fileName)
+        {
+            auto data = load_json_file(fileName);
+            return validate(data);
+        }
+        std::vector<std::string> validate(const json& data)
+        {
+            std::vector<std::string> errors;
+            // Validate the top-level object
+            validateObject("", schema, data, errors);
+            return errors;
+        }
+        json load_json_file(const std::string& fileName)
+        {
+            std::ifstream file(fileName);
+            if (!file.is_open()) {
+                throw std::runtime_error("Error: Unable to open file " + fileName);
+            }
+            json data;
+            file >> data;
+            file.close();
+            return data;
+        }
+        void fix_it(const std::string& fileName)
+        {
+            // Load JSON file
+            auto data = load_json_file(fileName);
+            // Fix fields
+            for (const auto& [key, value] : schema["properties"].items()) {
+                if (!data.contains(key)) {
+                    // Set default value if specified in the schema
+                    if (value.contains("default")) {
+                        data[key] = value["default"];
+                    } else if (value["type"] == "array") {
+                        data[key] = json::array();
+                    } else if (value["type"] == "object") {
+                        data[key] = json::object();
+                    } else {
+                        data[key] = nullptr;
+                    }
+                }
+                // Fix const fields to match the schema value
+                if (value.contains("const")) {
+                    data[key] = value["const"];
+                }
+            }
+            // Save fixed JSON
+            std::ofstream outFile(fileName);
+            if (!outFile.is_open()) {
+                std::cerr << "Error: Unable to open file for writing." << std::endl;
+                return;
+            }
+            outFile << data.dump(4);
+            outFile.close();
+        }
+
+    private:
+        json schema;
+
+        void validateObject(const std::string& path, const json& schema, const json& data, std::vector<std::string>& errors)
+        {
+            if (schema.contains("required")) {
+                for (const auto& requiredField : schema["required"]) {
+                    if (!data.contains(requiredField)) {
+                        std::string fullPath = path.empty() ? requiredField.get<std::string>() : path + "." + requiredField.get<std::string>();
+                        errors.push_back("Missing required field: " + fullPath);
+                    }
+                }
+            }
+            if (schema.contains("properties")) {
+                for (const auto& [key, value] : schema["properties"].items()) {
+                    if (data.contains(key)) {
+                        std::string fullPath = path.empty() ? key : path + "." + key;
+                        validateField(fullPath, value, data[key], errors);  // Pass data[key] for nested validation
+                    } else if (value.contains("required")) {
+                        errors.push_back("Missing required field: " + (path.empty() ? key : path + "." + key));
+                    }
+                }
+            }
+        }
+
+        void validateField(const std::string& field, const json& value, const json& data, std::vector<std::string>& errors)
+        {
+            if (value.contains("type")) {
+                const std::string& type = value["type"];
+                if (type == "array") {
+                    if (!data.is_array()) {
+                        errors.push_back("Field '" + field + "' should be an array.");
+                        return;
+                    }
+
+                    if (value.contains("items")) {
+                        for (size_t i = 0; i < data.size(); ++i) {
+                            validateObject(field + "[" + std::to_string(i) + "]", value["items"], data[i], errors);
+                        }
+                    }
+                } else if (type == "object") {
+                    if (!data.is_object()) {
+                        errors.push_back("Field '" + field + "' should be an object.");
+                        return;
+                    }
+
+                    validateObject(field, value, data, errors);
+                } else if (type == "string" && !data.is_string()) {
+                    errors.push_back("Field '" + field + "' should be a string.");
+                } else if (type == "number" && !data.is_number()) {
+                    errors.push_back("Field '" + field + "' should be a number.");
+                } else if (type == "integer" && !data.is_number_integer()) {
+                    errors.push_back("Field '" + field + "' should be an integer.");
+                } else if (type == "boolean" && !data.is_boolean()) {
+                    errors.push_back("Field '" + field + "' should be a boolean.");
+                }
+            }
+            if (value.contains("const")) {
+                const auto& expectedValue = value["const"];
+                if (data != expectedValue) {
+                    errors.push_back("Field '" + field + "' has an invalid value. Expected: " +
+                        expectedValue.dump() + ", Found: " + data.dump());
+                }
+            }
+        }
+    };
+}
+#endif

src/results/Result.cpp

@@ -8,6 +8,8 @@
 #include "common/Paths.h"
 #include "common/Symbols.h"
 #include "Result.h"
+#include "JsonValidator.h"
+#include "SchemaV1_0.h"
 
 namespace platform {
     std::string get_actual_date()
@@ -62,10 +64,14 @@ namespace platform {
     {
         return data;
     }
-
-    void Result::save()
+    std::vector<std::string> Result::check()
     {
-        std::ofstream file(Paths::results() + getFilename());
+        platform::JsonValidator validator(platform::SchemaV1_0::schema);
+        return validator.validate(data);
+    }
+    void Result::save(const std::string& path)
+    {
+        std::ofstream file(path + getFilename());
         file << data;
         file.close();
     }

src/results/Result.h

@@ -4,7 +4,7 @@
 #include <vector>
 #include <string>
 #include <nlohmann/json.hpp>
-#include "common/Timer.h"
+#include "common/Timer.hpp"
 #include "main/HyperParameters.h"
 #include "main/PartialResult.h"
 
@@ -15,7 +15,8 @@ namespace platform {
     public:
         Result();
         Result& load(const std::string& path, const std::string& filename);
-        void save();
+        void save(const std::string& path);
+        std::vector<std::string> check();
         // Getters
         json getJson();
         std::string to_string(int maxModel, int maxTitle) const;
@@ -28,7 +29,7 @@ namespace platform {
         std::string getModel() const { return data["model"].get<std::string>(); };
         std::string getPlatform() const { return data["platform"].get<std::string>(); };
         std::string getScoreName() const { return data["score_name"].get<std::string>(); };
-
+        void setSchemaVersion(const std::string& version) { data["schema_version"] = version; };
         bool isComplete() const { return complete; };
         json getData() const { return data; }
         // Setters

src/results/ResultsDatasetExcel.cpp

@@ -1,8 +1,9 @@
+#include "common/Paths.h"
 #include "ResultsDatasetExcel.h"
 namespace platform {
     ResultsDatasetExcel::ResultsDatasetExcel()
     {
-        file_name = "some_results.xlsx";
+        file_name = Paths::excelResults();
         workbook = workbook_new(getFileName().c_str());
         createFormats();
         setProperties("Results");

src/results/ResultsDatasetExcel.h

@@ -12,6 +12,7 @@ namespace platform {
         ResultsDatasetExcel();
         ~ResultsDatasetExcel();
         void report(json& data);
+        std::string getExcelFileName() { return getFileName(); }
     };
 }
 #endif

src/results/SchemaV1_0.h

@@ -0,0 +1,103 @@
+#ifndef SCHEMAV1_0_H
+#define SCHEMAV1_0_H
+#include <nlohmann/json.hpp>
+
+namespace platform {
+    using json = nlohmann::ordered_json;
+    class SchemaV1_0 {
+    public:
+        // Define JSON schema
+        const static json schema;
+
+    };
+    const json SchemaV1_0::schema = {
+        {"$schema", "http://json-schema.org/draft-07/schema#"},
+        {"type", "object"},
+        {"properties", {
+            {"schema_version", {
+                {"type", "string"},
+                {"pattern", "^\\d+\\.\\d+$"},
+                {"default", "1.0"},
+                {"const", "1.0"}  // Fixed schema version for this schema
+            }},
+            {"date", {{"type", "string"}, {"format", "date"}}},
+            {"time", {{"type", "string"}, {"pattern", "^\\d{2}:\\d{2}:\\d{2}$"}}},
+            {"title", {{"type", "string"}}},
+            {"language", {{"type", "string"}}},
+            {"language_version", {{"type", "string"}}},
+            {"discretized", {{"type", "boolean"}, {"default", false}}},
+            {"model", {{"type", "string"}}},
+            {"platform", {{"type", "string"}}},
+            {"stratified", {{"type", "boolean"}, {"default", false}}},
+            {"folds", {{"type", "integer"}, {"default", 0}}},
+            {"score_name", {{"type", "string"}}},
+            {"version", {{"type", "string"}}},
+            {"duration", {{"type", "number"}, {"default", 0}}},
+            {"results", {
+                {"type", "array"},
+                {"items", {
+                    {"type", "object"},
+                    {"properties", {
+                        {"scores_train", {{"type", "array"}, {"items", {{"type", "number"}}}}},
+                        {"scores_test", {{"type", "array"}, {"items", {{"type", "number"}}}}},
+                        {"times_train", {{"type", "array"}, {"items", {{"type", "number"}}}}},
+                        {"times_test", {{"type", "array"}, {"items", {{"type", "number"}}}}},
+                        {"notes", {{"type", "array"}, {"items", {{"type", "string"}}}}},
+                        {"train_time", {{"type", "number"}, {"default", 0}}},
+                        {"train_time_std", {{"type", "number"}, {"default", 0}}},
+                        {"test_time", {{"type", "number"}, {"default", 0}}},
+                        {"test_time_std", {{"type", "number"}, {"default", 0}}},
+                        {"samples", {{"type", "integer"}, {"default", 0}}},
+                        {"features", {{"type", "integer"}, {"default", 0}}},
+                        {"classes", {{"type", "integer"}, {"default", 0}}},
+                        {"hyperparameters", {
+                            {"type", "object"},
+                            {"additionalProperties", {
+                                {"oneOf", {
+                                    {{"type", "number"}},  // Field can be a number
+                                    {{"type", "string"}}   // Field can also be a string
+                                }}
+                            }}
+                        }},
+                        {"score", {{"type", "number"}, {"default", 0}}},
+                        {"score_train", {{"type", "number"}, {"default", 0}}},
+                        {"score_std", {{"type", "number"}, {"default", 0}}},
+                        {"score_train_std", {{"type", "number"}, {"default", 0}}},
+                        {"time", {{"type", "number"}, {"default", 0}}},
+                        {"time_std", {{"type", "number"}, {"default", 0}}},
+                        {"nodes", {{"type", "number"}, {"default", 0}}},
+                        {"leaves", {{"type", "number"}, {"default", 0}}},
+                        {"depth", {{"type", "number"}, {"default", 0}}},
+                        {"dataset", {{"type", "string"}}},
+                        {"confusion_matrices", {
+                            {"type", "array"},
+                            {"items", {
+                                {"type", "object"},
+                                {"patternProperties", {
+                                    {".*", {
+                                        {"type", "array"},
+                                        {"items", {{"type", "integer"}}}
+                                    }}
+                                }},
+                                {"additionalProperties", false}
+                            }}
+                        }}
+                    }},
+                    {"required", {
+                        "scores_train", "scores_test", "times_train", "times_test",
+                        "train_time", "train_time_std", "test_time", "test_time_std",
+                        "samples", "features", "classes", "hyperparameters", "score", "score_train",
+                        "score_std", "score_train_std", "time", "time_std", "nodes", "leaves",
+                        "depth", "dataset"
+                    }}
+                }}
+            }}
+        }},
+        {"required", {
+            "schema_version", "date", "time", "title", "language", "language_version",
+            "discretized", "model", "platform", "stratified", "folds", "score_name",
+            "version", "duration", "results"
+        }}
+    };
+}
+#endif

tests/CMakeLists.txt

@@ -2,21 +2,17 @@ if(ENABLE_TESTING)
     set(TEST_PLATFORM "unit_tests_platform")
     include_directories(
         ${Platform_SOURCE_DIR}/src
-        ${Platform_SOURCE_DIR}/lib/argparse/include
-        ${Platform_SOURCE_DIR}/lib/mdlp/src
-        ${Platform_SOURCE_DIR}/lib/Files
-        ${Platform_SOURCE_DIR}/lib/json/include
-        ${Platform_SOURCE_DIR}/lib/folding
         ${CMAKE_BINARY_DIR}/configured_files/include
-        ${PyClassifiers_INCLUDE_DIRS}
-        ${Bayesnet_INCLUDE_DIRS}
     )
     set(TEST_SOURCES_PLATFORM 
-        TestUtils.cpp TestPlatform.cpp TestResult.cpp TestScores.cpp 
+        TestUtils.cpp TestPlatform.cpp TestResult.cpp TestScores.cpp TestDecisionTree.cpp TestAdaBoost.cpp
         ${Platform_SOURCE_DIR}/src/common/Datasets.cpp ${Platform_SOURCE_DIR}/src/common/Dataset.cpp ${Platform_SOURCE_DIR}/src/common/Discretization.cpp
-        ${Platform_SOURCE_DIR}/src/main/Scores.cpp
+        ${Platform_SOURCE_DIR}/src/main/Scores.cpp 
+        ${Platform_SOURCE_DIR}/src/experimental_clfs/DecisionTree.cpp
+        ${Platform_SOURCE_DIR}/src/experimental_clfs/AdaBoost.cpp
     )
     add_executable(${TEST_PLATFORM} ${TEST_SOURCES_PLATFORM})
-    target_link_libraries(${TEST_PLATFORM} PUBLIC "${TORCH_LIBRARIES}" mdlp Catch2::Catch2WithMain BayesNet)
+    target_link_libraries(${TEST_PLATFORM} PUBLIC 
+      torch::torch fimdlp:fimdlp Catch2::Catch2WithMain bayesnet::bayesnet pyclassifiers::pyclassifiers)
     add_test(NAME ${TEST_PLATFORM} COMMAND ${TEST_PLATFORM})
 endif(ENABLE_TESTING)

tests/TestAdaBoost.cpp

@@ -0,0 +1,547 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include <catch2/catch_test_macros.hpp>
+#include <catch2/catch_approx.hpp>
+#include <catch2/matchers/catch_matchers_string.hpp>
+#include <catch2/matchers/catch_matchers_vector.hpp>
+#include <torch/torch.h>
+#include <memory>
+#include <stdexcept>
+#include "experimental_clfs/AdaBoost.h"
+#include "experimental_clfs/DecisionTree.h"
+#include "experimental_clfs/TensorUtils.hpp"
+#include "TestUtils.h"
+
+using namespace bayesnet;
+using namespace Catch::Matchers;
+
+static const bool DEBUG = false;
+
+TEST_CASE("AdaBoost Construction", "[AdaBoost]")
+{
+    SECTION("Default constructor")
+    {
+        REQUIRE_NOTHROW(AdaBoost());
+    }
+
+    SECTION("Constructor with parameters")
+    {
+        REQUIRE_NOTHROW(AdaBoost(100, 2));
+    }
+
+    SECTION("Constructor parameter access")
+    {
+        AdaBoost ada(75, 3);
+        REQUIRE(ada.getNEstimators() == 75);
+        REQUIRE(ada.getBaseMaxDepth() == 3);
+    }
+}
+
+TEST_CASE("AdaBoost Hyperparameter Setting", "[AdaBoost]")
+{
+    AdaBoost ada;
+
+    SECTION("Set individual hyperparameters")
+    {
+        REQUIRE_NOTHROW(ada.setNEstimators(100));
+        REQUIRE_NOTHROW(ada.setBaseMaxDepth(5));
+
+        REQUIRE(ada.getNEstimators() == 100);
+        REQUIRE(ada.getBaseMaxDepth() == 5);
+    }
+
+    SECTION("Set hyperparameters via JSON")
+    {
+        nlohmann::json params;
+        params["n_estimators"] = 80;
+        params["base_max_depth"] = 4;
+
+        REQUIRE_NOTHROW(ada.setHyperparameters(params));
+    }
+
+    SECTION("Invalid hyperparameters should throw")
+    {
+        nlohmann::json params;
+
+        // Negative n_estimators
+        params["n_estimators"] = -1;
+        REQUIRE_THROWS_AS(ada.setHyperparameters(params), std::invalid_argument);
+
+        // Zero n_estimators
+        params["n_estimators"] = 0;
+        REQUIRE_THROWS_AS(ada.setHyperparameters(params), std::invalid_argument);
+
+        // Negative base_max_depth
+        params["n_estimators"] = 50;
+        params["base_max_depth"] = -1;
+        REQUIRE_THROWS_AS(ada.setHyperparameters(params), std::invalid_argument);
+
+        // Zero base_max_depth
+        params["base_max_depth"] = 0;
+        REQUIRE_THROWS_AS(ada.setHyperparameters(params), std::invalid_argument);
+    }
+}
+
+TEST_CASE("AdaBoost Basic Functionality", "[AdaBoost]")
+{
+    // Create a simple dataset
+    int n_samples = 20;
+    int n_features = 2;
+
+    std::vector<std::vector<int>> X(n_features, std::vector<int>(n_samples));
+    std::vector<int> y(n_samples);
+
+    // Simple pattern: class depends on first feature
+    for (int i = 0; i < n_samples; i++) {
+        X[0][i] = i < 10 ? 0 : 1;
+        X[1][i] = i % 2;
+        y[i] = X[0][i];  // Class equals first feature
+    }
+
+    std::vector<std::string> features = { "f1", "f2" };
+    std::string className = "class";
+    std::map<std::string, std::vector<int>> states;
+    states["f1"] = { 0, 1 };
+    states["f2"] = { 0, 1 };
+    states["class"] = { 0, 1 };
+
+    SECTION("Training with vector interface")
+    {
+        AdaBoost ada(10, 3);  // 10 estimators, max_depth = 3
+        REQUIRE_NOTHROW(ada.fit(X, y, features, className, states, Smoothing_t::NONE));
+
+        // Check that we have the expected number of models
+        auto weights = ada.getEstimatorWeights();
+        REQUIRE(weights.size() <= 10);  // Should be <= n_estimators
+        REQUIRE(weights.size() > 0);    // Should have at least one model
+
+        // Check training errors
+        auto errors = ada.getTrainingErrors();
+        REQUIRE(errors.size() == weights.size());
+
+        // All training errors should be less than 0.5 for this simple dataset
+        for (double error : errors) {
+            REQUIRE(error < 0.5);
+            REQUIRE(error >= 0.0);
+        }
+    }
+
+    SECTION("Prediction before fitting")
+    {
+        AdaBoost ada;
+        REQUIRE_THROWS_WITH(ada.predict(X),
+            ContainsSubstring("not been fitted"));
+        REQUIRE_THROWS_WITH(ada.predict_proba(X),
+            ContainsSubstring("not been fitted"));
+    }
+
+    SECTION("Prediction with vector interface")
+    {
+        AdaBoost ada(10, 3);
+        ada.setDebug(DEBUG);  // Enable debug to investigate
+        ada.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        auto predictions = ada.predict(X);
+        REQUIRE(predictions.size() == static_cast<size_t>(n_samples));
+        // Check accuracy
+        int correct = 0;
+        for (size_t i = 0; i < predictions.size(); i++) {
+            if (predictions[i] == y[i]) correct++;
+        }
+        double accuracy = static_cast<double>(correct) / n_samples;
+        REQUIRE(accuracy > 0.99);  // Should achieve good accuracy on this simple dataset
+        auto accuracy_computed = ada.score(X, y);
+        REQUIRE(accuracy_computed == Catch::Approx(accuracy).epsilon(1e-6));
+    }
+
+    SECTION("Probability predictions with vector interface")
+    {
+        AdaBoost ada(10, 3);
+        ada.setDebug(DEBUG);  // ENABLE DEBUG HERE TOO
+        ada.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        auto proba = ada.predict_proba(X);
+        REQUIRE(proba.size() == static_cast<size_t>(n_samples));
+        REQUIRE(proba[0].size() == 2);  // Two classes
+
+        // Check probabilities sum to 1 and are valid
+        auto predictions = ada.predict(X);
+        int correct = 0;
+        for (size_t i = 0; i < proba.size(); i++) {
+            auto p = proba[i];
+            auto pred = predictions[i];
+            REQUIRE(p.size() == 2);
+            REQUIRE(p[0] >= 0.0);
+            REQUIRE(p[1] >= 0.0);
+            double sum = p[0] + p[1];
+            REQUIRE(sum == Catch::Approx(1.0).epsilon(1e-6));
+            // compute the predicted class based on probabilities
+            auto predicted_class = (p[0] > p[1]) ? 0 : 1;
+            // compute accuracy based on predictions
+            if (predicted_class == y[i]) {
+                correct++;
+            }
+
+            INFO("Probability test - Sample " << i << ": pred=" << pred << ", probs=[" << p[0] << "," << p[1] << "], expected_from_probs=" << predicted_class);
+
+            // Handle ties
+            if (std::abs(p[0] - p[1]) < 1e-10) {
+                INFO("Tie detected in probabilities");
+                // Either prediction is valid in case of tie
+            } else {
+                // Check that predict_proba matches the expected predict value
+                REQUIRE(pred == predicted_class);
+            }
+        }
+        double accuracy = static_cast<double>(correct) / n_samples;
+        REQUIRE(accuracy > 0.99);  // Should achieve good accuracy on this simple dataset
+    }
+}
+
+TEST_CASE("AdaBoost Tensor Interface", "[AdaBoost]")
+{
+    auto raw = RawDatasets("iris", true);
+
+    SECTION("Training with tensor format")
+    {
+        AdaBoost ada(20, 3);
+
+        INFO("Dataset shape: " << raw.dataset.sizes());
+        INFO("Features: " << raw.featurest.size());
+        INFO("Samples: " << raw.nSamples);
+
+        // AdaBoost expects dataset in format: features x samples, with labels as last row
+        REQUIRE_NOTHROW(ada.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, Smoothing_t::NONE));
+
+        // Test prediction with tensor
+        auto predictions = ada.predict(raw.Xt);
+        REQUIRE(predictions.size(0) == raw.yt.size(0));
+
+        // Calculate accuracy
+        auto correct = torch::sum(predictions == raw.yt).item<int>();
+        double accuracy = static_cast<double>(correct) / raw.yt.size(0);
+        auto accuracy_computed = ada.score(raw.Xt, raw.yt);
+        REQUIRE(accuracy_computed == Catch::Approx(accuracy).epsilon(1e-6));
+        REQUIRE(accuracy > 0.97);  // Should achieve good accuracy on Iris
+
+        // Test probability predictions with tensor
+        auto proba = ada.predict_proba(raw.Xt);
+        REQUIRE(proba.size(0) == raw.yt.size(0));
+        REQUIRE(proba.size(1) == 3);  // Three classes in Iris
+
+        // Check probabilities sum to 1
+        auto prob_sums = torch::sum(proba, 1);
+        for (int i = 0; i < prob_sums.size(0); i++) {
+            REQUIRE(prob_sums[i].item<double>() == Catch::Approx(1.0).epsilon(1e-6));
+        }
+    }
+}
+
+TEST_CASE("AdaBoost SAMME Algorithm Validation", "[AdaBoost]")
+{
+    auto raw = RawDatasets("iris", true);
+
+    SECTION("Prediction consistency with probabilities")
+    {
+        AdaBoost ada(15, 3);
+        ada.setDebug(DEBUG);  // Enable debug for ALL instances
+        ada.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, Smoothing_t::NONE);
+
+        auto predictions = ada.predict(raw.Xt);
+        auto probabilities = ada.predict_proba(raw.Xt);
+
+        REQUIRE(predictions.size(0) == probabilities.size(0));
+        REQUIRE(probabilities.size(1) == 3);  // Three classes in Iris
+
+        // For each sample, predicted class should correspond to highest probability
+        for (int i = 0; i < predictions.size(0); i++) {
+            int predicted_class = predictions[i].item<int>();
+            auto probs = probabilities[i];
+
+            // Find class with highest probability
+            auto max_prob_idx = torch::argmax(probs).item<int>();
+
+            // Predicted class should match class with highest probability
+            REQUIRE(predicted_class == max_prob_idx);
+
+            // Probabilities should sum to 1
+            double sum_probs = torch::sum(probs).item<double>();
+            REQUIRE(sum_probs == Catch::Approx(1.0).epsilon(1e-6));
+
+            // All probabilities should be non-negative
+            for (int j = 0; j < 3; j++) {
+                REQUIRE(probs[j].item<double>() >= 0.0);
+                REQUIRE(probs[j].item<double>() <= 1.0);
+            }
+        }
+    }
+
+    SECTION("Weighted voting verification")
+    {
+        // Simple dataset where we can verify the weighted voting
+        std::vector<std::vector<int>> X = { {0,0,1,1}, {0,1,0,1} };
+        std::vector<int> y = { 0, 1, 1, 0 };
+        std::vector<std::string> features = { "f1", "f2" };
+        std::string className = "class";
+        std::map<std::string, std::vector<int>> states;
+        states["f1"] = { 0, 1 };
+        states["f2"] = { 0, 1 };
+        states["class"] = { 0, 1 };
+
+        AdaBoost ada(5, 2);
+        ada.setDebug(DEBUG);  // Enable debug for detailed logging
+        ada.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        INFO("=== Final test verification ===");
+        auto predictions = ada.predict(X);
+        auto probabilities = ada.predict_proba(X);
+        auto alphas = ada.getEstimatorWeights();
+
+        INFO("Training info:");
+        for (size_t i = 0; i < alphas.size(); i++) {
+            INFO("  Model " << i << ": alpha=" << alphas[i]);
+        }
+
+        REQUIRE(predictions.size() == 4);
+        REQUIRE(probabilities.size() == 4);
+        REQUIRE(probabilities[0].size() == 2);  // Two classes
+        REQUIRE(alphas.size() > 0);
+
+        // Verify that estimator weights are reasonable
+        for (double alpha : alphas) {
+            REQUIRE(alpha >= 0.0);  // Alphas should be non-negative
+        }
+
+        // Verify prediction-probability consistency with detailed logging
+        for (size_t i = 0; i < predictions.size(); i++) {
+            int pred = predictions[i];
+            auto probs = probabilities[i];
+
+            INFO("Final check - Sample " << i << ": predicted=" << pred << ", probabilities=[" << probs[0] << "," << probs[1] << "]");
+
+            // Handle the case where probabilities are exactly equal (tie)
+            if (std::abs(probs[0] - probs[1]) < 1e-10) {
+                INFO("Tie detected in probabilities - either prediction is valid");
+                REQUIRE((pred == 0 || pred == 1));
+            } else {
+                // Normal case - prediction should match max probability
+                int expected_pred = (probs[0] > probs[1]) ? 0 : 1;
+                INFO("Expected prediction based on probs: " << expected_pred);
+                REQUIRE(pred == expected_pred);
+            }
+
+            REQUIRE(probs[0] + probs[1] == Catch::Approx(1.0).epsilon(1e-6));
+        }
+    }
+
+    SECTION("Empty models edge case")
+    {
+        AdaBoost ada(1, 1);
+        ada.setDebug(DEBUG);  // Enable debug for ALL instances
+
+        // Try to predict before fitting
+        std::vector<std::vector<int>> X = { {0}, {1} };
+        REQUIRE_THROWS_WITH(ada.predict(X), ContainsSubstring("not been fitted"));
+        REQUIRE_THROWS_WITH(ada.predict_proba(X), ContainsSubstring("not been fitted"));
+    }
+}
+
+TEST_CASE("AdaBoost Debug - Simple Dataset Analysis", "[AdaBoost][debug]")
+{
+    // Create the exact same simple dataset that was failing
+    int n_samples = 20;
+    int n_features = 2;
+
+    std::vector<std::vector<int>> X(n_features, std::vector<int>(n_samples));
+    std::vector<int> y(n_samples);
+
+    // Simple pattern: class depends on first feature
+    for (int i = 0; i < n_samples; i++) {
+        X[0][i] = i < 10 ? 0 : 1;
+        X[1][i] = i % 2;
+        y[i] = X[0][i];  // Class equals first feature
+    }
+
+    std::vector<std::string> features = { "f1", "f2" };
+    std::string className = "class";
+    std::map<std::string, std::vector<int>> states;
+    states["f1"] = { 0, 1 };
+    states["f2"] = { 0, 1 };
+    states["class"] = { 0, 1 };
+
+    SECTION("Debug training process")
+    {
+        AdaBoost ada(5, 3);  // Few estimators for debugging
+        ada.setDebug(DEBUG);
+
+        // This should work perfectly on this simple dataset
+        REQUIRE_NOTHROW(ada.fit(X, y, features, className, states, Smoothing_t::NONE));
+
+        // Get training details
+        auto weights = ada.getEstimatorWeights();
+        auto errors = ada.getTrainingErrors();
+
+        INFO("Number of models trained: " << weights.size());
+        INFO("Training errors: ");
+        for (size_t i = 0; i < errors.size(); i++) {
+            INFO("  Model " << i << ": error=" << errors[i] << ", weight=" << weights[i]);
+        }
+
+        // Should have at least one model
+        REQUIRE(weights.size() > 0);
+        REQUIRE(errors.size() == weights.size());
+
+        // All training errors should be reasonable for this simple dataset
+        for (double error : errors) {
+            REQUIRE(error >= 0.0);
+            REQUIRE(error < 0.5);  // Should be better than random
+        }
+
+        // Test predictions
+        auto predictions = ada.predict(X);
+        REQUIRE(predictions.size() == static_cast<size_t>(n_samples));
+
+        // Calculate accuracy
+        int correct = 0;
+        for (size_t i = 0; i < predictions.size(); i++) {
+            if (predictions[i] == y[i]) correct++;
+            INFO("Sample " << i << ": predicted=" << predictions[i] << ", actual=" << y[i]);
+        }
+        double accuracy = static_cast<double>(correct) / n_samples;
+        INFO("Accuracy: " << accuracy);
+
+        // Should achieve high accuracy on this perfectly separable dataset
+        REQUIRE(accuracy >= 0.9);  // Lower threshold for debugging
+
+        // Test probability predictions
+        auto proba = ada.predict_proba(X);
+        REQUIRE(proba.size() == static_cast<size_t>(n_samples));
+
+        // Verify probabilities are valid
+        for (size_t i = 0; i < proba.size(); i++) {
+            auto p = proba[i];
+            REQUIRE(p.size() == 2);
+            REQUIRE(p[0] >= 0.0);
+            REQUIRE(p[1] >= 0.0);
+            double sum = p[0] + p[1];
+            REQUIRE(sum == Catch::Approx(1.0).epsilon(1e-6));
+
+            // Predicted class should match highest probability
+            int pred_class = predictions[i];
+
+            // Handle ties
+            if (std::abs(p[0] - p[1]) < 1e-10) {
+                INFO("Tie detected - probabilities are equal");
+                REQUIRE((pred_class == 0 || pred_class == 1));
+            } else {
+                REQUIRE(pred_class == (p[0] > p[1] ? 0 : 1));
+            }
+        }
+    }
+
+    SECTION("Compare with single DecisionTree")
+    {
+        // Test that AdaBoost performs at least as well as a single tree
+        DecisionTree single_tree(3, 2, 1);
+        single_tree.fit(X, y, features, className, states, Smoothing_t::NONE);
+        auto tree_predictions = single_tree.predict(X);
+
+        int tree_correct = 0;
+        for (size_t i = 0; i < tree_predictions.size(); i++) {
+            if (tree_predictions[i] == y[i]) tree_correct++;
+        }
+        double tree_accuracy = static_cast<double>(tree_correct) / n_samples;
+
+        AdaBoost ada(5, 3);
+        ada.setDebug(DEBUG);
+        ada.fit(X, y, features, className, states, Smoothing_t::NONE);
+        auto ada_predictions = ada.predict(X);
+
+        int ada_correct = 0;
+        for (size_t i = 0; i < ada_predictions.size(); i++) {
+            if (ada_predictions[i] == y[i]) ada_correct++;
+        }
+        double ada_accuracy = static_cast<double>(ada_correct) / n_samples;
+
+        INFO("DecisionTree accuracy: " << tree_accuracy);
+        INFO("AdaBoost accuracy: " << ada_accuracy);
+
+        // AdaBoost should perform at least as well as single tree
+        // (allowing small tolerance for numerical differences)
+        REQUIRE(ada_accuracy >= tree_accuracy - 0.1);
+    }
+}
+
+TEST_CASE("AdaBoost Predict-Proba Consistency Fix", "[AdaBoost][consistency]")
+{
+    // Simple binary classification dataset
+    std::vector<std::vector<int>> X = { {0,0,1,1}, {0,1,0,1} };
+    std::vector<int> y = { 0, 0, 1, 1 };
+    std::vector<std::string> features = { "f1", "f2" };
+    std::string className = "class";
+    std::map<std::string, std::vector<int>> states;
+    states["f1"] = { 0, 1 };
+    states["f2"] = { 0, 1 };
+    states["class"] = { 0, 1 };
+
+    SECTION("Binary classification consistency")
+    {
+        AdaBoost ada(3, 2);
+        ada.setDebug(DEBUG);  // Enable debug output
+        ada.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        INFO("=== Debugging predict vs predict_proba consistency ===");
+
+        // Get training info
+        auto alphas = ada.getEstimatorWeights();
+        auto errors = ada.getTrainingErrors();
+
+        INFO("Training completed:");
+        INFO("  Number of models: " << alphas.size());
+        for (size_t i = 0; i < alphas.size(); i++) {
+            INFO("  Model " << i << ": alpha=" << alphas[i] << ", error=" << errors[i]);
+        }
+
+        auto predictions = ada.predict(X);
+        auto probabilities = ada.predict_proba(X);
+
+        // Verify consistency for each sample
+        for (size_t i = 0; i < predictions.size(); i++) {
+            int predicted_class = predictions[i];
+            auto probs = probabilities[i];
+
+            INFO("Sample " << i << ":");
+            INFO("  Features: [" << X[0][i] << ", " << X[1][i] << "]");
+            INFO("  True class: " << y[i]);
+            INFO("  Predicted class: " << predicted_class);
+            INFO("  Probabilities: [" << probs[0] << ", " << probs[1] << "]");
+
+            // The predicted class should be the one with highest probability
+            int max_prob_class = (probs[0] > probs[1]) ? 0 : 1;
+            INFO("  Max prob class: " << max_prob_class);
+
+            // Handle tie case (when probabilities are equal)
+            if (std::abs(probs[0] - probs[1]) < 1e-10) {
+                INFO("  Tie detected - probabilities are equal");
+                // In case of tie, either prediction is valid
+                REQUIRE((predicted_class == 0 || predicted_class == 1));
+            } else {
+                REQUIRE(predicted_class == max_prob_class);
+            }
+
+            // Probabilities should sum to 1
+            double sum_probs = probs[0] + probs[1];
+            REQUIRE(sum_probs == Catch::Approx(1.0).epsilon(1e-6));
+
+            // All probabilities should be valid
+            REQUIRE(probs[0] >= 0.0);
+            REQUIRE(probs[1] >= 0.0);
+            REQUIRE(probs[0] <= 1.0);
+            REQUIRE(probs[1] <= 1.0);
+        }
+    }
+}

tests/TestDecisionTree.cpp

@@ -0,0 +1,311 @@
+// ***************************************************************
+// SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
+// SPDX-FileType: SOURCE
+// SPDX-License-Identifier: MIT
+// ***************************************************************
+
+#include <catch2/catch_test_macros.hpp>
+#include <catch2/catch_approx.hpp>
+#include <catch2/matchers/catch_matchers_string.hpp>
+#include <catch2/matchers/catch_matchers_vector.hpp>
+#include <torch/torch.h>
+#include <memory>
+#include <stdexcept>
+#include "experimental_clfs/DecisionTree.h"
+#include "TestUtils.h"
+
+using namespace bayesnet;
+using namespace Catch::Matchers;
+
+TEST_CASE("DecisionTree Construction", "[DecisionTree]")
+{
+    SECTION("Default constructor")
+    {
+        REQUIRE_NOTHROW(DecisionTree());
+    }
+
+    SECTION("Constructor with parameters")
+    {
+        REQUIRE_NOTHROW(DecisionTree(5, 10, 3));
+    }
+}
+
+TEST_CASE("DecisionTree Hyperparameter Setting", "[DecisionTree]")
+{
+    DecisionTree dt;
+
+    SECTION("Set individual hyperparameters")
+    {
+        REQUIRE_NOTHROW(dt.setMaxDepth(10));
+        REQUIRE_NOTHROW(dt.setMinSamplesSplit(5));
+        REQUIRE_NOTHROW(dt.setMinSamplesLeaf(2));
+        REQUIRE(dt.getMaxDepth() == 10);
+        REQUIRE(dt.getMinSamplesSplit() == 5);
+        REQUIRE(dt.getMinSamplesLeaf() == 2);
+    }
+
+    SECTION("Set hyperparameters via JSON")
+    {
+        nlohmann::json params;
+        params["max_depth"] = 7;
+        params["min_samples_split"] = 4;
+        params["min_samples_leaf"] = 2;
+
+        REQUIRE_NOTHROW(dt.setHyperparameters(params));
+        REQUIRE(dt.getMaxDepth() == 7);
+        REQUIRE(dt.getMinSamplesSplit() == 4);
+        REQUIRE(dt.getMinSamplesLeaf() == 2);
+    }
+
+    SECTION("Invalid hyperparameters should throw")
+    {
+        nlohmann::json params;
+
+        // Negative max_depth
+        params["max_depth"] = -1;
+        REQUIRE_THROWS_AS(dt.setHyperparameters(params), std::invalid_argument);
+
+        // Zero min_samples_split
+        params["max_depth"] = 5;
+        params["min_samples_split"] = 0;
+        REQUIRE_THROWS_AS(dt.setHyperparameters(params), std::invalid_argument);
+
+        // Negative min_samples_leaf
+        params["min_samples_split"] = 2;
+        params["min_samples_leaf"] = -5;
+        REQUIRE_THROWS_AS(dt.setHyperparameters(params), std::invalid_argument);
+    }
+}
+
+TEST_CASE("DecisionTree Basic Functionality", "[DecisionTree]")
+{
+    // Create a simple dataset
+    int n_samples = 20;
+    int n_features = 2;
+
+    std::vector<std::vector<int>> X(n_features, std::vector<int>(n_samples));
+    std::vector<int> y(n_samples);
+
+    // Simple pattern: class depends on first feature
+    for (int i = 0; i < n_samples; i++) {
+        X[0][i] = i < 10 ? 0 : 1;
+        X[1][i] = i % 2;
+        y[i] = X[0][i];  // Class equals first feature
+    }
+
+    std::vector<std::string> features = { "f1", "f2" };
+    std::string className = "class";
+    std::map<std::string, std::vector<int>> states;
+    states["f1"] = { 0, 1 };
+    states["f2"] = { 0, 1 };
+    states["class"] = { 0, 1 };
+
+    SECTION("Training with vector interface")
+    {
+        DecisionTree dt(3, 2, 1);
+        REQUIRE_NOTHROW(dt.fit(X, y, features, className, states, Smoothing_t::NONE));
+
+        auto predictions = dt.predict(X);
+        REQUIRE(predictions.size() == static_cast<size_t>(n_samples));
+
+        // Should achieve perfect accuracy on this simple dataset
+        int correct = 0;
+        for (size_t i = 0; i < predictions.size(); i++) {
+            if (predictions[i] == y[i]) correct++;
+        }
+        REQUIRE(correct == n_samples);
+    }
+
+    SECTION("Prediction before fitting")
+    {
+        DecisionTree dt;
+        REQUIRE_THROWS_WITH(dt.predict(X),
+            ContainsSubstring("Classifier has not been fitted"));
+    }
+
+    SECTION("Probability predictions")
+    {
+        DecisionTree dt(3, 2, 1);
+        dt.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        auto proba = dt.predict_proba(X);
+        REQUIRE(proba.size() == static_cast<size_t>(n_samples));
+        REQUIRE(proba[0].size() == 2);  // Two classes
+
+        // Check probabilities sum to 1 and probabilities are valid
+        auto predictions = dt.predict(X);
+        for (size_t i = 0; i < proba.size(); i++) {
+            auto p = proba[i];
+            auto pred = predictions[i];
+            REQUIRE(p.size() == 2);
+            REQUIRE(p[0] >= 0.0);
+            REQUIRE(p[1] >= 0.0);
+            double sum = p[0] + p[1];
+            //Check that prodict_proba matches the expected predict value
+            REQUIRE(pred == (p[0] > p[1] ? 0 : 1));
+            REQUIRE(sum == Catch::Approx(1.0).epsilon(1e-6));
+        }
+    }
+}
+
+TEST_CASE("DecisionTree on Iris Dataset", "[DecisionTree][iris]")
+{
+    auto raw = RawDatasets("iris", true);
+
+    SECTION("Training with dataset format")
+    {
+        DecisionTree dt(5, 2, 1);
+
+        INFO("Dataset shape: " << raw.dataset.sizes());
+        INFO("Features: " << raw.featurest.size());
+        INFO("Samples: " << raw.nSamples);
+
+        // DecisionTree expects dataset in format: features x samples, with labels as last row
+        REQUIRE_NOTHROW(dt.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, Smoothing_t::NONE));
+
+        // Test prediction
+        auto predictions = dt.predict(raw.Xt);
+        REQUIRE(predictions.size(0) == raw.yt.size(0));
+
+        // Calculate accuracy
+        auto correct = torch::sum(predictions == raw.yt).item<int>();
+        double accuracy = static_cast<double>(correct) / raw.yt.size(0);
+        double acurracy_computed = dt.score(raw.Xt, raw.yt);
+        REQUIRE(accuracy > 0.97);  // Reasonable accuracy for Iris
+        REQUIRE(acurracy_computed == Catch::Approx(accuracy).epsilon(1e-6));
+    }
+
+    SECTION("Training with vector interface")
+    {
+        DecisionTree dt(5, 2, 1);
+
+        REQUIRE_NOTHROW(dt.fit(raw.Xv, raw.yv, raw.featuresv, raw.classNamev, raw.statesv, Smoothing_t::NONE));
+
+        // std::cout << "Tree structure:\n";
+        // auto graph_lines = dt.graph("Iris Decision Tree");
+        // for (const auto& line : graph_lines) {
+        //     std::cout << line << "\n";
+        // }
+        auto predictions = dt.predict(raw.Xv);
+        REQUIRE(predictions.size() == raw.yv.size());
+    }
+
+    SECTION("Different tree depths")
+    {
+        std::vector<int> depths = { 1, 3, 5 };
+
+        for (int depth : depths) {
+            DecisionTree dt(depth, 2, 1);
+            dt.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, Smoothing_t::NONE);
+
+            auto predictions = dt.predict(raw.Xt);
+            REQUIRE(predictions.size(0) == raw.yt.size(0));
+        }
+    }
+}
+
+TEST_CASE("DecisionTree Edge Cases", "[DecisionTree]")
+{
+    auto raw = RawDatasets("iris", true);
+
+    SECTION("Very shallow tree")
+    {
+        DecisionTree dt(1, 2, 1);  // depth = 1
+        dt.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, Smoothing_t::NONE);
+
+        auto predictions = dt.predict(raw.Xt);
+        REQUIRE(predictions.size(0) == raw.yt.size(0));
+
+        // With depth 1, should have at most 2 unique predictions
+        auto unique_vals = at::_unique(predictions);
+        REQUIRE(std::get<0>(unique_vals).size(0) <= 2);
+    }
+
+    SECTION("High min_samples_split")
+    {
+        DecisionTree dt(10, 50, 1);
+        dt.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, Smoothing_t::NONE);
+
+        auto predictions = dt.predict(raw.Xt);
+        REQUIRE(predictions.size(0) == raw.yt.size(0));
+    }
+}
+
+TEST_CASE("DecisionTree Graph Visualization", "[DecisionTree]")
+{
+    // Simple dataset
+    std::vector<std::vector<int>> X = { {0,0,0,1}, {0,1,1,1} };  // XOR pattern
+    std::vector<int> y = { 0, 1, 1, 0 };  // XOR pattern
+    std::vector<std::string> features = { "x1", "x2" };
+    std::string className = "xor";
+    std::map<std::string, std::vector<int>> states;
+    states["x1"] = { 0, 1 };
+    states["x2"] = { 0, 1 };
+    states["xor"] = { 0, 1 };
+
+    SECTION("Graph generation")
+    {
+        DecisionTree dt(2, 1, 1);
+        dt.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        auto graph_lines = dt.graph();
+
+        REQUIRE(graph_lines.size() > 2);
+        REQUIRE(graph_lines.front() == "digraph DecisionTree {");
+        REQUIRE(graph_lines.back() == "}");
+
+        // Should contain node definitions
+        bool has_nodes = false;
+        for (const auto& line : graph_lines) {
+            if (line.find("node") != std::string::npos) {
+                has_nodes = true;
+                break;
+            }
+        }
+        REQUIRE(has_nodes);
+    }
+
+    SECTION("Graph with title")
+    {
+        DecisionTree dt(2, 1, 1);
+        dt.fit(X, y, features, className, states, Smoothing_t::NONE);
+
+        auto graph_lines = dt.graph("XOR Tree");
+
+        bool has_title = false;
+        for (const auto& line : graph_lines) {
+            if (line.find("label=\"XOR Tree\"") != std::string::npos) {
+                has_title = true;
+                break;
+            }
+        }
+        REQUIRE(has_title);
+    }
+}
+
+TEST_CASE("DecisionTree with Weights", "[DecisionTree]")
+{
+    auto raw = RawDatasets("iris", true);
+
+    SECTION("Uniform weights")
+    {
+        DecisionTree dt(5, 2, 1);
+        dt.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, raw.weights, Smoothing_t::NONE);
+
+        auto predictions = dt.predict(raw.Xt);
+        REQUIRE(predictions.size(0) == raw.yt.size(0));
+    }
+
+    SECTION("Non-uniform weights")
+    {
+        auto weights = torch::ones({ raw.nSamples });
+        weights.index({ torch::indexing::Slice(0, 50) }) *= 2.0;  // Emphasize first class
+        weights = weights / weights.sum();
+
+        DecisionTree dt(5, 2, 1);
+        dt.fit(raw.dataset, raw.featurest, raw.classNamet, raw.statest, weights, Smoothing_t::NONE);
+
+        auto predictions = dt.predict(raw.Xt);
+        REQUIRE(predictions.size(0) == raw.yt.size(0));
+    }
+}

tests/TestPlatform.cpp

@@ -20,17 +20,17 @@ TEST_CASE("Test Platform version", "[Platform]")
 TEST_CASE("Test Folding library version", "[Folding]")
 {
     std::string version = folding::KFold(5, 100).version();
-    REQUIRE(version == "1.1.0");
+    REQUIRE(version == "1.1.1");
 }
 TEST_CASE("Test BayesNet version", "[BayesNet]")
 {
     std::string version = bayesnet::TAN().getVersion();
-    REQUIRE(version == "1.0.6");
+    REQUIRE(version == "1.1.2");
 }
 TEST_CASE("Test mdlp version", "[mdlp]")
 {
     std::string version = mdlp::CPPFImdlp::version();
-    REQUIRE(version == "2.0.0");
+    REQUIRE(version == "2.0.1");
 }
 TEST_CASE("Test Arff version", "[Arff]")
 {

tests/TestScores.cpp

@@ -14,38 +14,40 @@
 using json = nlohmann::ordered_json;
 auto epsilon = 1e-4;
 
-void make_test_bin(int TP, int TN, int FP, int FN, std::vector<int>& y_test, std::vector<int>& y_pred)
+void make_test_bin(int TP, int TN, int FP, int FN, std::vector<int>& y_test, torch::Tensor& y_pred)
 {
-    // TP
+    std::vector<std::array<double, 2>> probs;
+    // TP: true positive (label 1, predicted 1)
     for (int i = 0; i < TP; i++) {
         y_test.push_back(1);
-        y_pred.push_back(1);
+        probs.push_back({ 0.0, 1.0 }); // P(class 0)=0, P(class 1)=1
     }
-    // TN
+    // TN: true negative (label 0, predicted 0)
     for (int i = 0; i < TN; i++) {
         y_test.push_back(0);
-        y_pred.push_back(0);
+        probs.push_back({ 1.0, 0.0 }); // P(class 0)=1, P(class 1)=0
     }
-    // FP
+    // FP: false positive (label 0, predicted 1)
     for (int i = 0; i < FP; i++) {
         y_test.push_back(0);
-        y_pred.push_back(1);
+        probs.push_back({ 0.0, 1.0 }); // P(class 0)=0, P(class 1)=1
     }
-    // FN
+    // FN: false negative (label 1, predicted 0)
     for (int i = 0; i < FN; i++) {
         y_test.push_back(1);
-        y_pred.push_back(0);
+        probs.push_back({ 1.0, 0.0 }); // P(class 0)=1, P(class 1)=0
     }
+    // Convert to torch::Tensor of double, shape [N,2]
+    y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 2 }, torch::kFloat64).clone();
 }
 
 TEST_CASE("Scores binary", "[Scores]")
 {
     std::vector<int> y_test;
-    std::vector<int> y_pred;
+    torch::Tensor y_pred;
     make_test_bin(197, 210, 52, 41, y_test, y_pred);
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 2);
+    platform::Scores scores(y_test_tensor, y_pred, 2);
     REQUIRE(scores.accuracy() == Catch::Approx(0.814).epsilon(epsilon));
     REQUIRE(scores.f1_score(0) == Catch::Approx(0.818713));
     REQUIRE(scores.f1_score(1) == Catch::Approx(0.809035));
@@ -64,10 +66,23 @@ TEST_CASE("Scores binary", "[Scores]")
 TEST_CASE("Scores multiclass", "[Scores]")
 {
     std::vector<int> y_test = { 0, 2, 2, 2, 2, 0, 1, 2, 0, 2 };
-    std::vector<int> y_pred = { 0, 1, 2, 2, 1, 1, 1, 0, 0, 2 };
+    // Refactor y_pred to a tensor of shape [10, 3] with probabilities
+    std::vector<std::array<double, 3>> probs = {
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 0.0, 1.0 } // P(class 0)=0, P(class 1)=0, P(class 2)=1
+    };
+    torch::Tensor y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 3 }, torch::kFloat64).clone();
+    // Convert y_test to a tensor
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 3);
+    platform::Scores scores(y_test_tensor, y_pred, 3);
     REQUIRE(scores.accuracy() == Catch::Approx(0.6).epsilon(epsilon));
     REQUIRE(scores.f1_score(0) == Catch::Approx(0.666667));
     REQUIRE(scores.f1_score(1) == Catch::Approx(0.4));
@@ -84,10 +99,21 @@ TEST_CASE("Scores multiclass", "[Scores]")
 TEST_CASE("Test Confusion Matrix Values", "[Scores]")
 {
     std::vector<int> y_test = { 0, 2, 2, 2, 2, 0, 1, 2, 0, 2 };
-    std::vector<int> y_pred = { 0, 1, 2, 2, 1, 1, 1, 0, 0, 2 };
+    std::vector<std::array<double, 3>> probs = {
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 0.0, 1.0 } // P(class 0)=0, P(class 1)=0, P(class 2)=1
+    };
+    torch::Tensor y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 3 }, torch::kFloat64).clone();
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 3);
+    platform::Scores scores(y_test_tensor, y_pred, 3);
     auto confusion_matrix = scores.get_confusion_matrix();
     REQUIRE(confusion_matrix[0][0].item<int>() == 2);
     REQUIRE(confusion_matrix[0][1].item<int>() == 1);
@@ -102,11 +128,22 @@ TEST_CASE("Test Confusion Matrix Values", "[Scores]")
 TEST_CASE("Confusion Matrix JSON", "[Scores]")
 {
     std::vector<int> y_test = { 0, 2, 2, 2, 2, 0, 1, 2, 0, 2 };
-    std::vector<int> y_pred = { 0, 1, 2, 2, 1, 1, 1, 0, 0, 2 };
+    std::vector<std::array<double, 3>> probs = {
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 0.0, 1.0 } // P(class 0)=0, P(class 1)=0, P(class 2)=1
+    };
+    torch::Tensor y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 3 }, torch::kFloat64).clone();
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
     std::vector<std::string> labels = { "Aeroplane", "Boat", "Car" };
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 3, labels);
+    platform::Scores scores(y_test_tensor, y_pred, 3, labels);
     auto res_json_int = scores.get_confusion_matrix_json();
     REQUIRE(res_json_int[0][0] == 2);
     REQUIRE(res_json_int[0][1] == 1);
@@ -131,11 +168,22 @@ TEST_CASE("Confusion Matrix JSON", "[Scores]")
 TEST_CASE("Classification Report", "[Scores]")
 {
     std::vector<int> y_test = { 0, 2, 2, 2, 2, 0, 1, 2, 0, 2 };
-    std::vector<int> y_pred = { 0, 1, 2, 2, 1, 1, 1, 0, 0, 2 };
+    std::vector<std::array<double, 3>> probs = {
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 0.0, 1.0 } // P(class 0)=0, P(class 1)=0, P(class 2)=1
+    };
+    torch::Tensor y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 3 }, torch::kFloat64).clone();
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
     std::vector<std::string> labels = { "Aeroplane", "Boat", "Car" };
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 3, labels);
+    platform::Scores scores(y_test_tensor, y_pred, 3, labels);
     auto report = scores.classification_report(Colors::BLUE(), "train");
     auto json_matrix = scores.get_confusion_matrix_json(true);
     platform::Scores scores2(json_matrix);
@@ -144,11 +192,22 @@ TEST_CASE("Classification Report", "[Scores]")
 TEST_CASE("JSON constructor", "[Scores]")
 {
     std::vector<int> y_test = { 0, 2, 2, 2, 2, 0, 1, 2, 0, 2 };
-    std::vector<int> y_pred = { 0, 1, 2, 2, 1, 1, 1, 0, 0, 2 };
+    std::vector<std::array<double, 3>> probs = {
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 0.0, 1.0 } // P(class 0)=0, P(class 1)=0, P(class 2)=1
+    };
+    torch::Tensor y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 3 }, torch::kFloat64).clone();
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
     std::vector<std::string> labels = { "Car", "Boat", "Aeroplane" };
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 3, labels);
+    platform::Scores scores(y_test_tensor, y_pred, 3, labels);
     auto res_json_int = scores.get_confusion_matrix_json();
     platform::Scores scores2(res_json_int);
     REQUIRE(scores.accuracy() == scores2.accuracy());
@@ -173,17 +232,14 @@ TEST_CASE("JSON constructor", "[Scores]")
 TEST_CASE("Aggregate", "[Scores]")
 {
     std::vector<int> y_test;
-    std::vector<int> y_pred;
+    torch::Tensor y_pred;
     make_test_bin(197, 210, 52, 41, y_test, y_pred);
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 2);
+    platform::Scores scores(y_test_tensor, y_pred, 2);
     y_test.clear();
-    y_pred.clear();
     make_test_bin(227, 187, 39, 47, y_test, y_pred);
     auto y_test_tensor2 = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor2 = torch::tensor(y_pred, torch::kInt32);
-    platform::Scores scores2(y_test_tensor2, y_pred_tensor2, 2);
+    platform::Scores scores2(y_test_tensor2, y_pred, 2);
     scores.aggregate(scores2);
     REQUIRE(scores.accuracy() == Catch::Approx(0.821).epsilon(epsilon));
     REQUIRE(scores.f1_score(0) == Catch::Approx(0.8160329));
@@ -195,11 +251,9 @@ TEST_CASE("Aggregate", "[Scores]")
     REQUIRE(scores.f1_weighted() == Catch::Approx(0.8209856));
     REQUIRE(scores.f1_macro() == Catch::Approx(0.8208694));
     y_test.clear();
-    y_pred.clear();
     make_test_bin(197 + 227, 210 + 187, 52 + 39, 41 + 47, y_test, y_pred);
     y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
-    platform::Scores scores3(y_test_tensor, y_pred_tensor, 2);
+    platform::Scores scores3(y_test_tensor, y_pred, 2);
     for (int i = 0; i < 2; ++i) {
         REQUIRE(scores3.f1_score(i) == scores.f1_score(i));
         REQUIRE(scores3.precision(i) == scores.precision(i));
@@ -212,11 +266,22 @@ TEST_CASE("Aggregate", "[Scores]")
 TEST_CASE("Order of keys", "[Scores]")
 {
     std::vector<int> y_test = { 0, 2, 2, 2, 2, 0, 1, 2, 0, 2 };
-    std::vector<int> y_pred = { 0, 1, 2, 2, 1, 1, 1, 0, 0, 2 };
+    std::vector<std::array<double, 3>> probs = {
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 0.0, 1.0 }, // P(class 0)=0, P(class 1)=0, P(class 2)=1
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 0.0, 1.0, 0.0 }, // P(class 0)=0, P(class 1)=1, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 1.0, 0.0, 0.0 }, // P(class 0)=1, P(class 1)=0, P(class 2)=0
+        { 0.0, 0.0, 1.0 } // P(class 0)=0, P(class 1)=0, P(class 2)=1
+    };
+    torch::Tensor y_pred = torch::from_blob(probs.data(), { (long)probs.size(), 3 }, torch::kFloat64).clone();
     auto y_test_tensor = torch::tensor(y_test, torch::kInt32);
-    auto y_pred_tensor = torch::tensor(y_pred, torch::kInt32);
     std::vector<std::string> labels = { "Car", "Boat", "Aeroplane" };
-    platform::Scores scores(y_test_tensor, y_pred_tensor, 3, labels);
+    platform::Scores scores(y_test_tensor, y_pred, 3, labels);
     auto res_json_int = scores.get_confusion_matrix_json(true);
     // Make a temp file and store the json
     std::string filename = "temp.json";

tests/TestUtils.h

@@ -5,7 +5,7 @@
 #include <vector>
 #include <map>
 #include <tuple>
-#include <ArffFiles.hpp>
+#include <ArffFiles/ArffFiles.hpp>
 #include <fimdlp/CPPFImdlp.h>
 
 bool file_exists(const std::string& name);