rmontanana/Platform
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add log and fix some mistakes in integration

Browse Source
This commit is contained in:
Ricardo Montañana Gómez
2025-02-25 20:35:13 +01:00
parent de7cf091be
commit c63baf419f
5 changed files with 3573 additions and 251 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2009
lib/log/loguru.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

1475
lib/log/loguru.hpp Normal file
View File

File diff suppressed because it is too large Load Diff

1
src/CMakeLists.txt
View File

@@ -1,5 +1,6 @@
include_directories(
## Libs
${Platform_SOURCE_DIR}/lib/log
${Platform_SOURCE_DIR}/lib/Files
${Platform_SOURCE_DIR}/lib/folding
${Platform_SOURCE_DIR}/lib/mdlp/src

81
src/experimental_clfs/XBAODE.cpp
View File

@@ -10,10 +10,14 @@
#include <tuple>
#include "XBAODE.h"
#include "TensorUtils.hpp"
#include <loguru.hpp>
#include <loguru.cpp>
namespace platform {
XBAODE::XBAODE() : semaphore_{ CountingSemaphore::getInstance() }, Boost(false)
{
validHyperparameters = { "alpha_block", "order", "convergence", "convergence_best", "bisection", "threshold", "maxTolerance",
"predict_voting", "select_features" };
}
void XBAODE::trainModel(const torch::Tensor& weights, const bayesnet::Smoothing_t smoothing)
{
@@ -22,23 +26,23 @@ namespace platform {
y_train_ = TensorUtils::to_vector<int>(y_train);
X_test_ = TensorUtils::to_matrix(X_test);
y_test_ = TensorUtils::to_vector<int>(y_test);
maxTolerance = 5;
//
// Logging setup
//
// loguru::set_thread_name("XBAODE");
// loguru::g_stderr_verbosity = loguru::Verbosity_OFF;
// loguru::add_file("boostAODE.log", loguru::Truncate, loguru::Verbosity_MAX);
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;
torch::Tensor weights_ = torch::full({ m }, 1.0, torch::kFloat64);
torch::Tensor weights_ = torch::full({ m }, 1.0 / m, torch::kFloat64);
bool finished = false;
std::vector<int> featuresUsed;
int num_instances = m;
int num_attributes = n;
significanceModels.resize(num_attributes, 0.0);
significanceModels.resize(n, 0.0); // n possible spodes
aode_.fit(X_train_, y_train_, features, className, states, smoothing);
n_models = 0;
if (selectFeatures) {
featuresUsed = featureSelection(weights_);
aode_.set_active_parents(featuresUsed);
@@ -49,6 +53,8 @@ namespace platform {
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;
}
@@ -78,46 +84,41 @@ namespace platform {
);
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());
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());
aode_.add_active_parent(feature);
alpha_t = 0.0;
if (!block_update) {
std::vector<int> ypred;
if (alpha_block) {
//
// Compute the prediction with the current ensemble + model
//
// Add the model to the ensemble
n_models++;
significanceModels[feature] = 1.0;
aode_.add_active_parent(feature);
// Compute the prediction
ypred = aode_.predict(X_train_);
// Remove the model from the ensemble
significanceModels[feature] = 0.0;
aode_.remove_last_parent();
n_models--;
} else {
ypred = aode_.predict_spode(X_train_, feature);
}
// 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_);
std::vector<int> ypred;
if (alpha_block) {
//
// Compute the prediction with the current ensemble + model
//
// Add the model to the ensemble
n_models++;
significanceModels[feature] = 1.0;
aode_.add_active_parent(feature);
// Compute the prediction
ypred = aode_.predict(X_train_);
// Remove the model from the ensemble
significanceModels[feature] = 0.0;
aode_.remove_last_parent();
n_models--;
} else {
ypred = aode_.predict_spode(X_train_, feature);
}
// 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);
aode_.add_active_parent(feature);
significanceModels.push_back(alpha_t);
n_models++;
// VLOG_SCOPE_F(2, "numItemsPack: %d n_models: %d featuresUsed: %zu", numItemsPack, n_models, featuresUsed.size());
}
if (block_update) {
std::tie(weights_, alpha_t, finished) = update_weights_block(k, y_train, weights_);
}
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 = predict(X_test);
double accuracy = (y_val_predict == y_test).sum().item<double>() / (double)y_test.size(0);
@@ -127,10 +128,10 @@ namespace platform {
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);
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);
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;
}
@@ -142,13 +143,13 @@ namespace platform {
priorAccuracy = accuracy;
}
}
// VLOG_SCOPE_F(1, "tolerance: %d featuresUsed.size: %zu features.size: %zu", tolerance, featuresUsed.size(), features.size());
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);
VLOG_SCOPE_F(4, "Convergence threshold reached & %d models eliminated of %d", numItemsPack, n_models);
for (int i = 0; i < numItemsPack; ++i) {
significanceModels.pop_back();
models.pop_back();
@@ -156,7 +157,7 @@ namespace platform {
}
} 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);
VLOG_SCOPE_F(4, "Convergence threshold reached & 0 models eliminated n_models=%d numItemsPack=%d", n_models, numItemsPack);
}
}
if (featuresUsed.size() != features.size()) {

258
src/experimental_clfs/Xaode.hpp
View File

@@ -40,9 +40,8 @@ namespace platform {
// where states.back() = number of class states.
//
// We'll store:
// 1) p(c) in classPriors_
// 2) p(x_i=si | c) in classFeatureProbs_
// 3) p(x_j=sj | c, x_i=si) in data_, with i<j => i is "superparent," j is "child."
// 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.
@@ -98,9 +97,8 @@ namespace platform {
classFeatureCounts_.resize(feature_offset * statesClass_);
classFeatureProbs_.resize(feature_offset * statesClass_);
// classCounts_[c] & p(c) in classPriors_
// classCounts_[c]
classCounts_.resize(statesClass_, 0.0);
classPriors_.resize(statesClass_, 0.0);
matrixState_ = MatrixState::COUNTS;
}
@@ -122,8 +120,6 @@ namespace platform {
for (int s : states_) std::cout << s << " "; std::cout << std::endl;
std::cout << "- classCounts: size: " << classCounts_.size() << std::endl;
for (double cc : classCounts_) std::cout << cc << " "; std::cout << std::endl;
std::cout << "- classPriors: size: " << classPriors_.size() << std::endl;
for (double cp : classPriors_) std::cout << cp << " "; std::cout << std::endl;
std::cout << "- classFeatureCounts: size: " << classFeatureCounts_.size() << std::endl;
for (double cfc : classFeatureCounts_) std::cout << cfc << " "; std::cout << std::endl;
std::cout << "- classFeatureProbs: size: " << classFeatureProbs_.size() << std::endl;
@@ -191,29 +187,16 @@ namespace platform {
// -------------------------------------------------------
//
// Once all samples are added in COUNTS mode, call this to:
// 1) compute class priors p(c)
// 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)
// 1) compute p(x_i=si | c) => classFeatureProbs_
// 2) 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.");
}
// (1) p(c)
double totalCount = std::accumulate(classCounts_.begin(), classCounts_.end(), 0.0);
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] / totalCount;
}
}
// (2) p(x_i=si | c) => classFeatureProbs_
// (1) p(x_i=si | c) => classFeatureProbs_
int idx, sf;
double denom, countVal, p;
for (int feature = 0; feature < nFeatures_; ++feature) {
@@ -237,8 +220,8 @@ namespace platform {
}
}
// getCountFromTable(int classVal, int pIndex, int childIndex)
// (3) p(x_j=sj | c, x_i=si) => data_(i,si,j,sj,c)
// (3) p(x_i=si | c, x_j=sj) => dataOpp_(j,sj,i,si,c)
// (2) p(x_j=sj | c, x_i=si) => data_(i,si,j,sj,c)
// (2) p(x_i=si | c, x_j=sj) => dataOpp_(j,sj,i,si,c)
double pccCount, pcCount, ccCount;
double conditionalProb, oppositeCondProb;
int part1, part2, p1, part2_class, p1_class;
@@ -286,76 +269,66 @@ namespace platform {
// 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) const
std::vector<double> predict_proba_spode(const std::vector<int>& instance, int parent)
{
if (matrixState_ != MatrixState::PROBS) {
throw std::logic_error("predict_proba_spode: Xaode not in PROBS state.");
}
if ((int)instance.size() != nFeatures_) {
throw std::invalid_argument("predict_proba_spode: instance.size() != nFeatures_.");
}
if (parent < 0 || parent >= nFeatures_) {
throw std::out_of_range("predict_proba_spode: invalid superparent index.");
}
std::vector<double> scores(statesClass_, 0.0);
// accumulates posterior probabilities for each class
auto probs = std::vector<double>(statesClass_);
auto spodeProbs = std::vector<double>(statesClass_);
// Initialize the probabilities with the feature|class probabilities
int localOffset;
int sp = instance[parent];
int idx;
double pSpGivenC, pChildGivenSp, product;
double base;
double offset = (featureClassOffset_[parent] + sp) * statesClass_;
double parent_offset = pairOffset_[featureClassOffset_[parent] + sp];
// For each class c
localOffset = (featureClassOffset_[parent] + sp) * statesClass_;
for (int c = 0; c < statesClass_; ++c) {
// Start with p(c) * p(x_sp=spState| c)
pSpGivenC = classFeatureProbs_[offset + c];
product = pSpGivenC;
bool zeroProb = false;
for (int feature = 0; feature < nFeatures_; ++feature) {
if (feature == parent) continue;
int sf = instance[feature];
// Retrieve p(x_i= state_i | c, x_sp= spState)
base = (parent_offset + featureClassOffset_[feature] + sf) * statesClass_;
idx = base + c;
pChildGivenSp = data_[idx] * dataOpp_[idx];
if (pChildGivenSp <= 0.0) {
zeroProb = true;
break;
}
product *= pChildGivenSp;
}
scores[c] = zeroProb ? 0.0 : product;
spodeProbs[c] = classFeatureProbs_[localOffset + c];
}
normalize(scores);
return scores;
int idx, base, sc, parent_offset;
sp = instance[parent];
parent_offset = pairOffset_[featureClassOffset_[parent] + sp];
for (int child = 0; child < parent; ++child) {
sc = instance[child];
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_
*/
/*
int base = pairOffset_[i * nFeatures_ + j];
int blockSize = states_[i] * states_[j];
return base + c * blockSize + (si * states_[j] + sj);
*/
// index = compute_index(parent, instance[parent], child, instance[child], classVal);
idx = base + c;
spodeProbs[c] *= data_[idx];
spodeProbs[c] *= dataOpp_[idx];
}
}
// Normalize the probabilities
normalize(probs);
return probs;
}
int predict_spode(const std::vector<int>& instance, int parent) const
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()));
}
std::vector<double> predict_proba(std::vector<int>& instance)
std::vector<double> predict_proba(const std::vector<int>& instance)
{
Timer timer;
timer.start();
if (matrixState_ != MatrixState::PROBS) {
throw std::logic_error("predict_proba: Xaode not in PROBS state.");
}
if ((int)instance.size() != nFeatures_) {
throw std::invalid_argument("predict_proba: instance.size() != nFeatures_.");
}
// 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];
}
}
duration_first += timer.getDuration(); timer.start();
int idx, base, sp, sc, parent_offset;
for (int parent = 1; parent < nFeatures_; ++parent) {
// if parent is not in the active_parents, skip it
@@ -386,7 +359,6 @@ namespace platform {
}
}
}
duration_second += timer.getDuration(); timer.start();
/* add all the probabilities for each class */
for (int c = 0; c < statesClass_; ++c) {
for (int i = 0; i < nFeatures_; ++i) {
@@ -414,140 +386,6 @@ namespace platform {
}
}
// -------------------------------------------------------
// checkCoherence
// -------------------------------------------------------
//
// Check that the class priors, feature–class distributions and pairwise conditionals
// are coherent. They have to sum to 1.0 within a threshold.
//
void checkCoherenceApprox(double threshold) const
{
if (matrixState_ != MatrixState::PROBS) {
throw std::logic_error("checkCoherenceApprox: must be in PROBS state.");
}
// ------------------------------------------------------------------
// 1) Check that sum of class priors ~ 1
// ------------------------------------------------------------------
double sumPriors = 0.0;
for (double pc : classPriors_) {
sumPriors += pc;
}
if (std::fabs(sumPriors - 1.0) > threshold) {
std::ostringstream oss;
oss << "Xaode::checkCoherenceApprox - sum of classPriors = " << sumPriors
<< ", differs from 1.0 by more than " << threshold;
throw std::runtime_error(oss.str());
}
// ------------------------------------------------------------------
// 2) For each feature i and class c, the sum over all states si of
// classFeatureProbs_ should match the prior p(c) ~ classPriors_[c].
//
// (Because if you're storing p(x_i=si, c)/total or a scaled version,
// summing over si is effectively p(c).)
// ------------------------------------------------------------------
for (int c = 0; c < statesClass_; ++c) {
for (int i = 0; i < nFeatures_; ++i) {
double sumFeature = 0.0;
for (int si = 0; si < states_[i]; ++si) {
int idx = (featureClassOffset_[i] + si) * statesClass_ + c;
sumFeature += classFeatureProbs_[idx];
}
double expected = classPriors_[c];
if (std::fabs(sumFeature - expected) > threshold) {
std::ostringstream oss;
oss << "Xaode::checkCoherenceApprox - sum_{si} classFeatureProbs_ "
<< "for (feature=" << i << ", class=" << c << ") = " << sumFeature
<< ", expected ~ " << expected
<< ", difference is " << std::fabs(sumFeature - expected)
<< " > threshold=" << threshold;
throw std::runtime_error(oss.str());
}
}
}
// ------------------------------------------------------------------
// 3) For data_: sum_{child states} data_ should match the "parent" row
// in classFeatureProbs_, i.e. p(x_i=si, c).
//
// Because if data_[... i, si, j, sj, c] holds something like
// p(x_i=si, x_j=sj, c) (or a scaled fraction),
// then sum_{ sj } data_ = p(x_i=si, c).
// ------------------------------------------------------------------
for (int parent = 1; parent < nFeatures_; ++parent) {
for (int child = 0; child < parent; ++child) {
for (int c = 0; c < statesClass_; ++c) {
for (int spVal = 0; spVal < states_[parent]; ++spVal) {
double sumChildProb = 0.0;
// pairOffset_ gives the offset for (parent featureVal),
// then we add the child's offset and multiply by statesClass_.
int part1 = pairOffset_[featureClassOffset_[parent] + spVal];
for (int scVal = 0; scVal < states_[child]; ++scVal) {
int part2 = featureClassOffset_[child] + scVal;
int idx = (part1 + part2) * statesClass_ + c;
sumChildProb += data_[idx];
}
// Compare with classFeatureProbs_[parent, spVal, c]
double expected = classFeatureProbs_[
(featureClassOffset_[parent] + spVal) * statesClass_ + c
];
if (std::fabs(sumChildProb - expected) > threshold) {
std::ostringstream oss;
oss << "Xaode::checkCoherenceApprox - sum_{sj} data_ "
<< "for (parentFeature=" << parent
<< ", parentVal=" << spVal
<< ", childFeature=" << child
<< ", class=" << c << ") = " << sumChildProb
<< ", expected ~ " << expected
<< ", diff " << std::fabs(sumChildProb - expected)
<< " > threshold=" << threshold;
throw std::runtime_error(oss.str());
}
}
}
}
}
// ------------------------------------------------------------------
// 4) For dataOpp_: sum_{parent states} dataOpp_ should match the "child"
// row in classFeatureProbs_, i.e. p(x_j=sj, c).
// ------------------------------------------------------------------
for (int parent = 1; parent < nFeatures_; ++parent) {
for (int child = 0; child < parent; ++child) {
for (int c = 0; c < statesClass_; ++c) {
for (int scVal = 0; scVal < states_[child]; ++scVal) {
double sumParentProb = 0.0;
int part2 = featureClassOffset_[child] + scVal;
for (int spVal = 0; spVal < states_[parent]; ++spVal) {
int part1 = pairOffset_[featureClassOffset_[parent] + spVal];
int idx = (part1 + part2) * statesClass_ + c;
sumParentProb += dataOpp_[idx];
}
// Compare with classFeatureProbs_[child, scVal, c]
double expected = classFeatureProbs_[
(featureClassOffset_[child] + scVal) * statesClass_ + c
];
if (std::fabs(sumParentProb - expected) > threshold) {
std::ostringstream oss;
oss << "Xaode::checkCoherenceApprox - sum_{spVal} dataOpp_ "
<< "for (childFeature=" << child
<< ", childVal=" << scVal
<< ", parentFeature=" << parent
<< ", class=" << c << ") = " << sumParentProb
<< ", expected ~ " << expected
<< ", diff " << std::fabs(sumParentProb - expected)
<< " > threshold=" << threshold;
throw std::runtime_error(oss.str());
}
}
}
}
}
// If we get here, all sums are coherent under this "joint distribution" interpretation
}
int statesClass() const
{
return statesClass_;
@@ -602,8 +440,6 @@ namespace platform {
std::vector<double> classFeatureCounts_;
std::vector<double> classFeatureProbs_; // => p(x_i=si | c) after normalization
std::vector<double> classPriors_; // => p(c)
MatrixState matrixState_;
double SMOOTHING = 1.0;