Impelement CutPoints 2 with points of view

fimdlp/CPPFImdlp.cpp

@@ -1,4 +1,6 @@
 #include "CPPFImdlp.h"
+#include <numeric>
+#include <iostream>
 namespace CPPFImdlp
 {
     CPPFImdlp::CPPFImdlp()
@@ -7,23 +9,38 @@ namespace CPPFImdlp
     CPPFImdlp::~CPPFImdlp()
     {
     }
-    std::vector<float> CPPFImdlp::cutPoints(std::vector<int> &X, std::vector<int> &y)
+    std::vector<double> CPPFImdlp::cutPoints(std::vector<float> &X, std::vector<int> &y)
     {
-        std::vector<float> cutPts;
+        std::vector<double> cutPts;
-        int i, ant = X.at(0), anty = y.at(0);
+        double antx;
-        int n = X.size();
+        // int anty;
-        for (i = 1; i < n; i++)
+        std::vector<size_t> indices = sortIndices(X);
+        antx = X.at(indices[0]);
+        // anty = y.at(indices[0]);
+        for (auto index = indices.begin(); index != indices.end(); ++index)
         {
-            if (X.at(i) != ant)
+            // std::cout << X.at(*index) << " -> " << y.at(*index) << " // ";
+            //  Definition 2 Cut points are always on boundaries
+            // if (y.at(*index) != anty && antx < X.at(*index))
+            //  Weka implementation
+            if (antx < X.at(*index))
             {
-                if (y.at(i) != anty)
+                // std::cout << "* (" << X.at(*index) << ", " << antx << ") // ";
-                {
+                cutPts.push_back((X.at(*index) + antx) / 2);
-                    cutPts.push_back(float(X.at(i) + ant) / 2);
+                // anty = y.at(*index);
-                    ant = X.at(i);
-                    anty = y.at(i);
-                }
             }
+            antx = X.at(*index);
         }
+        // std::cout << std::endl;
         return cutPts;
     }
+    std::vector<size_t> CPPFImdlp::sortIndices(std::vector<float> &X)
+    {
+        std::vector<size_t> idx(X.size());
+        std::iota(idx.begin(), idx.end(), 0);
+        for (std::size_t i = 0; i < X.size(); i++)
+            stable_sort(idx.begin(), idx.end(), [&X](size_t i1, size_t i2)
+                        { return X[i1] < X[i2]; });
+        return idx;
+    }
 }

fimdlp/CPPFImdlp.h

@@ -2,14 +2,18 @@
 #define CPPFIMDLP_H
 #include <vector>
 #include <Python.h>
+#include <utility>
 namespace CPPFImdlp
 {
     class CPPFImdlp
     {
+    private:
+        std::vector<size_t> sortIndices(std::vector<float> &);
+
     public:
         CPPFImdlp();
         ~CPPFImdlp();
-        std::vector<float> cutPoints(std::vector<int> &, std::vector<int> &);
+        std::vector<double> cutPoints(std::vector<float> &, std::vector<int> &);
     };
 }
 #endif

fimdlp/cfimdlp.pyx

@@ -5,7 +5,7 @@ from libcpp.vector cimport vector
 cdef extern from "CPPFImdlp.h" namespace "CPPFImdlp":
     cdef cppclass CPPFImdlp:
         CPPFImdlp() except + 
-        vector[float] cutPoints(vector[int]&, vector[int]&)
+        vector[double] cutPoints(vector[float]&, vector[int]&)
 
 cdef class CFImdlp:
     cdef CPPFImdlp *thisptr

fimdlp/mdlp.py

@@ -96,8 +96,17 @@ class FImdlp(TransformerMixin, BaseEstimator):
                 "Shape of input is different from what was seen" "in `fit`"
             )
         print("Cut points for each feature in Iris dataset:")
-        for i in range(0, self.n_features_):
+        yz = self.y_.copy()
-            data = np.sort(X[:, i])
+        xz = X[:, 0].copy()
-            Xcutpoints = self.discretizer_.cut_points(data, self.y_)
+        xzz = self.discretizer_.sort_vectors(xz, yz)
+        print("Xz: ", xz)
+        print("Yz: ", yz)
+        print("Xzz: ", xzz)
+        print("Solución:")
+        print("Xz*: ", np.sort(X[:, 0]))
+        print("yz*: ", yz[np.argsort(X[:, 0])])
+        for i in range(0, 1):  # self.n_features_):
+            datax = np.sort(X[:, i])
+            Xcutpoints = self.discretizer_.cut_points(datax, self.y_)
             print(f"{self.features_[i]:20s}: {Xcutpoints}")
         return X

sample.py

@@ -1,9 +1,13 @@
 from sklearn.datasets import load_iris
 from fimdlp.mdlp import FImdlp
+from fimdlp.cppfimdlp import CFImdlp
 
 data = load_iris()
 X = data.data
 y = data.target
 features = data.feature_names
-test = FImdlp()
+# test = FImdlp()
-Xcutpoints = test.fit(X, y, features=features).transform(X)
+# Xcutpoints = test.fit(X, y, features=features).transform(X)
+clf = CFImdlp()
+print("Cut points for feature 0 in Iris dataset:")
+print(clf.cut_points(X[:, 0], y))

setup.py

@@ -10,9 +10,13 @@ setup(
     ext_modules=[
         Extension(
             name="cppfimdlp",
-            sources=["fimdlp/cfimdlp.pyx", "fimdlp/CPPFImdlp.cpp"],
+            sources=[
+                "fimdlp/cfimdlp.pyx",
+                "fimdlp/CPPFImdlp.cpp",
+            ],
             language="c++",
             include_dirs=["fimdlp"],
+            extra_compile_args=["-std=c++20"],
         ),
     ]
 )