AdaBoost.cpp

/*
GRT MIT License
Copyright (c) <2012> <Nicholas Gillian, Media Lab, MIT>

Permission is hereby granted, free of charge, to any person obtaining a copy of this software
and associated documentation files (the "Software"), to deal in the Software without restriction,
including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense,
and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so,
subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial
portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT
LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/

#define GRT_DLL_EXPORTS
#include "AdaBoost.h"

GRT_BEGIN_NAMESPACE

//Define the string that will be used to identify the object
const std::string AdaBoost::id = "AdaBoost";
std::string AdaBoost::getId() { return AdaBoost::id; }

//Register the AdaBoost module with the Classifier base class
RegisterClassifierModule< AdaBoost > AdaBoost::registerModule( getId() );

AdaBoost::AdaBoost(const WeakClassifier &weakClassifier,bool useScaling,bool useNullRejection,Float nullRejectionCoeff,UINT numBoostingIterations,UINT predictionMethod) : Classifier( AdaBoost::getId() )
{
 setWeakClassifier( weakClassifier );
 this->useScaling = useScaling;
 this->useNullRejection = useNullRejection;
 this->nullRejectionCoeff = nullRejectionCoeff;
 this->numBoostingIterations = numBoostingIterations;
 this->predictionMethod = predictionMethod;
 classifierMode = STANDARD_CLASSIFIER_MODE;
}

AdaBoost::AdaBoost(const AdaBoost &rhs) : Classifier( AdaBoost::getId() )
{
 classifierMode = STANDARD_CLASSIFIER_MODE;
 *this = rhs;
}

AdaBoost::~AdaBoost(void)
{
 //Clean up any weak classifiers
 clearWeakClassifiers();
}

AdaBoost& AdaBoost::operator=(const AdaBoost &rhs){
 if( this != &rhs ){
 //Clear the current weak classifiers
 clearWeakClassifiers();
 
 this->numBoostingIterations = rhs.numBoostingIterations;
 this->predictionMethod = rhs.predictionMethod;
 this->models = rhs.models;
 
 if( rhs.weakClassifiers.getSize() > 0 ){
 this->weakClassifiers.reserve( rhs.weakClassifiers.getSize() );
 for(UINT i=0; i<rhs.weakClassifiers.getSize(); i++){
 WeakClassifier *weakClassiferPtr = rhs.weakClassifiers[i]->createNewInstance();
 weakClassifiers.push_back( weakClassiferPtr );
 }
 }
 
 //Clone the classifier variables
 copyBaseVariables( dynamic_cast<const Classifier*>( &rhs ) );
 }
 return *this;
}

bool AdaBoost::deepCopyFrom(const Classifier *classifier){
 
 if( classifier == NULL ){
 errorLog << "deepCopyFrom(const Classifier *classifier) - The classifier pointer is NULL!" << std::endl;
 return false;
 }
 
 if( this->getId() == classifier->getId() ){
 //Clone the AdaBoost values
 const AdaBoost *ptr = dynamic_cast<const AdaBoost*>( classifier );
 
 //Clear the current weak classifiers
 clearWeakClassifiers();
 
 this->numBoostingIterations = ptr->numBoostingIterations;
 this->predictionMethod = ptr->predictionMethod;
 this->models = ptr->models;
 
 if( ptr->weakClassifiers.size() > 0 ){
 this->weakClassifiers.resize( ptr->weakClassifiers.getSize() );
 for(UINT i=0; i<ptr->weakClassifiers.getSize(); i++){
 weakClassifiers[i] = ptr->weakClassifiers[i]->createNewInstance();
 }
 }
 
 //Clone the classifier variables
 return copyBaseVariables( classifier );
 }
 return false;
}

bool AdaBoost::train_(ClassificationData &trainingData){
 
 //Clear any previous model
 clear();
 
 if( trainingData.getNumSamples() <= 1 ){
 errorLog << "train_(ClassificationData &trainingData) - There are not enough training samples to train a model! Number of samples: " << trainingData.getNumSamples() << std::endl;
 return false;
 }

 numInputDimensions = trainingData.getNumDimensions();
 numOutputDimensions = trainingData.getNumClasses();
 numClasses = trainingData.getNumClasses();
 const UINT POSITIVE_LABEL = WEAK_CLASSIFIER_POSITIVE_CLASS_LABEL;
 const UINT NEGATIVE_LABEL = WEAK_CLASSIFIER_NEGATIVE_CLASS_LABEL;
 Float alpha = 0;
 const Float beta = 0.001;
 Float epsilon = 0;
 TrainingResult trainingResult;
 ClassificationData validationData;
 
 const UINT K = weakClassifiers.getSize();
 if( K == 0 ){
 errorLog << "train_(ClassificationData &trainingData) - No weakClassifiers have been set. You need to set at least one weak classifier first." << std::endl;
 return false;
 }

 //Pass the logging state onto the weak classifiers
 for(UINT k=0; k<K; k++){
 weakClassifiers[k]->setTrainingLoggingEnabled( this->getTrainingLoggingEnabled() );
 }
 
 classLabels.resize(numClasses);
 models.resize(numClasses);
 ranges = trainingData.getRanges();
 
 //Scale the training data if needed
 if( useScaling ){
 trainingData.scale(ranges,0,1);
 }

 if( useValidationSet ){
 validationData = trainingData.split( 100-validationSetSize );
 }

 const UINT M = trainingData.getNumSamples();
 trainingLog << "Training AdaBoost model, num training examples: " << M << ", num validation examples: " << validationData.getNumSamples() << ", num classes: " << numClasses << ", num weak learners: " << K << std::endl;

 //Create the weights vector
 VectorFloat weights(M);
 
 //Create the error matrix
 MatrixFloat errorMatrix(K,M);
 
 for(UINT classIter=0; classIter<numClasses; classIter++){
 
 //Get the class label for the current class
 classLabels[classIter] = trainingData.getClassLabels()[classIter];
 
 //Set the class label of the current model
 models[ classIter ].setClassLabel( classLabels[classIter] );
 
 //Setup the labels for this class, POSITIVE_LABEL == 1, NEGATIVE_LABEL == 2
 ClassificationData classData;
 classData.setNumDimensions(trainingData.getNumDimensions());
 for(UINT i=0; i<M; i++){
 UINT label = trainingData[i].getClassLabel()==classLabels[classIter] ? POSITIVE_LABEL : NEGATIVE_LABEL;
 VectorFloat trainingSample = trainingData[i].getSample();
 classData.addSample(label,trainingSample);
 }
 
 //Setup the initial training sample weights
 std::fill(weights.begin(),weights.end(),1.0/M);
 
 //Run the boosting loop
 bool keepBoosting = true;
 UINT t = 0;
 
 while( keepBoosting ){
 
 //Pick the classifier from the family of classifiers that minimizes the total error
 UINT bestClassifierIndex = 0;
 Float minError = grt_numeric_limits< Float >::max();
 for(UINT k=0; k<K; k++){
 //Get the k'th possible classifier
 WeakClassifier *weakLearner = weakClassifiers[k];
 
 //Train the current classifier
 if( !weakLearner->train(classData,weights) ){
 errorLog << __GRT_LOG__ << " Failed to train weakLearner!" << std::endl;
 return false;
 }
 
 //Compute the weighted error for this clasifier
 Float e = 0;
 Float positiveLabel = weakLearner->getPositiveClassLabel();
 Float numCorrect = 0;
 Float numIncorrect = 0;
 for(UINT i=0; i<M; i++){
 //Only penalize errors
 Float prediction = weakLearner->predict( classData[i].getSample() );
 
 if( (prediction == positiveLabel && classData[i].getClassLabel() != POSITIVE_LABEL) || //False positive
 (prediction != positiveLabel && classData[i].getClassLabel() == POSITIVE_LABEL) ){ //False negative
 e += weights[i]; //Increase the error proportional to the weight of the example
 errorMatrix[k][i] = 1; //Flag that there was an error
 numIncorrect++;
 }else{
 errorMatrix[k][i] = 0; //Flag that there was no error
 numCorrect++;
 }
 }
 
 trainingLog << "PositiveClass: " << classLabels[classIter] << " Boosting Iter: " << t << " Classifier: " << k << " WeightedError: " << e << " NumCorrect: " << numCorrect/M << " NumIncorrect: " <<numIncorrect/M << std::endl;
 
 if( e < minError ){
 minError = e;
 bestClassifierIndex = k;
 }
 
 }
 
 epsilon = minError;
 
 //Set alpha, using the M1 weight value, small weights (close to 0) will receive a strong weight in the final classifier
 alpha = 0.5 * log( (1.0-epsilon)/epsilon );
 
 trainingLog << "PositiveClass: " << classLabels[classIter] << " Boosting Iter: " << t << " Best Classifier Index: " << bestClassifierIndex << " MinError: " << minError << " Alpha: " << alpha << std::endl;
 
 if( grt_isinf(alpha) ){ keepBoosting = false; trainingLog << "Alpha is INF. Stopping boosting for current class" << std::endl; }
 if( 0.5 - epsilon <= beta ){ keepBoosting = false; trainingLog << "Epsilon <= Beta. Stopping boosting for current class" << std::endl; }
 if( ++t >= numBoostingIterations ) keepBoosting = false;
 
 trainingResult.setClassificationResult(t, minError, this);
 trainingResults.push_back(trainingResult);
 trainingResultsObserverManager.notifyObservers( trainingResult );
 
 if( keepBoosting ){
 
 //Add the best weak classifier to the committee
 models[ classIter ].addClassifierToCommitee( weakClassifiers[bestClassifierIndex], alpha );
 
 //Update the weights for the next boosting iteration
 Float reWeight = (1.0 - epsilon) / epsilon;
 Float oldSum = 0;
 Float newSum = 0;
 for(UINT i=0; i<M; i++){
 oldSum += weights[i];
 //Only update the weights that resulted in an incorrect prediction
 if( errorMatrix[bestClassifierIndex][i] == 1 ) weights[i] *= reWeight;
 newSum += weights[i];
 }
 
 //Normalize all the weights
 //This results to increasing the weights of the samples that were incorrectly labelled
 //While decreasing the weights of the samples that were correctly classified
 reWeight = oldSum/newSum;
 for(UINT i=0; i<M; i++){
 weights[i] *= reWeight;
 }
 
 }else{
 trainingLog << "Stopping boosting training at iteration : " << t-1 << " with an error of " << epsilon << std::endl;
 if( t-1 == 0 ){
 //Add the best weak classifier to the committee (we have to add it as this is the first iteration)
 if( grt_isinf(alpha) ){ alpha = 1; } //If alpha is infinite then the first classifier got everything correct
 models[ classIter ].addClassifierToCommitee( weakClassifiers[bestClassifierIndex], alpha );
 }
 }
 
 }
 }
 
 //Normalize the weights
 for(UINT k=0; k<numClasses; k++){
 models[k].normalizeWeights();
 }
 
 //Flag that the model has been fully trained
 trained = true;
 converged = true;
 
 //Setup the data for prediction
 predictedClassLabel = 0;
 maxLikelihood = 0;
 classLikelihoods.resize(numClasses);
 classDistances.resize(numClasses);

 //Compute the final training stats
 trainingSetAccuracy = 0;
 validationSetAccuracy = 0;

 //If scaling was on, then the data will already be scaled, so turn it off temporially
 bool scalingState = useScaling;
 useScaling = false;
 for(UINT i=0; i<M; i++){
 if( !predict_( trainingData[i].getSample() ) ){
 trained = false;
 errorLog << __GRT_LOG__ << " Failed to run prediction for training sample: " << i << "! Failed to fully train model!" << std::endl;
 return false;
 }

 if( predictedClassLabel == trainingData[i].getClassLabel() ){
 trainingSetAccuracy++;
 }
 }

 if( useValidationSet ){
 for(UINT i=0; i<validationData.getNumSamples(); i++){
 if( !predict_( validationData[i].getSample() ) ){
 trained = false;
 errorLog << __GRT_LOG__ << " Failed to run prediction for validation sample: " << i << "! Failed to fully train model!" << std::endl;
 return false;
 }

 if( predictedClassLabel == validationData[i].getClassLabel() ){
 validationSetAccuracy++;
 }
 }
 }

 trainingSetAccuracy = trainingSetAccuracy / M * 100.0;

 trainingLog << "Training set accuracy: " << trainingSetAccuracy << std::endl;

 if( useValidationSet ){
 validationSetAccuracy = validationSetAccuracy / validationData.getNumSamples() * 100.0;
 trainingLog << "Validation set accuracy: " << validationSetAccuracy << std::endl;
 }

 //Reset the scaling state for future prediction
 useScaling = scalingState;

 return true;
}

bool AdaBoost::predict_(VectorFloat &inputVector){
 
 predictedClassLabel = 0;
 maxLikelihood = -10000;
 
 if( !trained ){
 errorLog << __GRT_LOG__ << " AdaBoost Model Not Trained!" << std::endl;
 return false;
 }
 
 if( inputVector.getSize() != numInputDimensions ){
 errorLog << __GRT_LOG__ << " The size of the input vector (" << inputVector.size() << ") does not match the num features in the model (" << numInputDimensions << std::endl;
 return false;
 }
 
 if( useScaling ){
 for(UINT n=0; n<numInputDimensions; n++){
 inputVector[n] = scale(inputVector[n], ranges[n].minValue, ranges[n].maxValue, 0, 1);
 }
 }
 
 if( classLikelihoods.getSize() != numClasses ) classLikelihoods.resize(numClasses,0);
 if( classDistances.getSize() != numClasses ) classDistances.resize(numClasses,0);
 
 UINT bestClassIndex = 0;
 UINT numPositivePredictions = 0;
 bestDistance = -grt_numeric_limits< Float >::max();
 Float worstDistance = grt_numeric_limits< Float >::max();
 Float sum = 0;
 for(UINT k=0; k<numClasses; k++){
 Float result = models[k].predict( inputVector );
 
 switch ( predictionMethod ) {
 case MAX_POSITIVE_VALUE:
 if( result > 0 ){
 if( result > bestDistance ){
 bestDistance = result;
 bestClassIndex = k;
 }
 numPositivePredictions++;
 classLikelihoods[k] = result;
 }else classLikelihoods[k] = 0;
 
 classDistances[k] = result;
 sum += classLikelihoods[k];
 
 break;
 case MAX_VALUE:
 if( result > bestDistance ){
 bestDistance = result;
 bestClassIndex = k;
 }
 if( result < worstDistance ){
 worstDistance = result;
 }
 numPositivePredictions++; //In the MAX_VALUE mode we assume all samples are valid
 classLikelihoods[k] = result;
 classDistances[k] = result;
 
 break;
 default:
 errorLog << __GRT_LOG__ << " Unknown prediction method!" << std::endl;
 break;
 }
 }
 
 if( predictionMethod == MAX_VALUE ){
 //Some of the class likelihoods might be negative, so we add the most negative value to each to offset this
 worstDistance = fabs( worstDistance );
 for(UINT k=0; k<numClasses; k++){
 classLikelihoods[k] += worstDistance;
 sum += classLikelihoods[k];
 }
 }
 
 //Normalize the class likelihoods
 if( sum > 0 ){
 for(UINT k=0; k<numClasses; k++)
 classLikelihoods[k] /= sum;
 }
 maxLikelihood = classLikelihoods[ bestClassIndex ];
 
 if( numPositivePredictions == 0 ){
 predictedClassLabel = GRT_DEFAULT_NULL_CLASS_LABEL;
 }else predictedClassLabel = classLabels[ bestClassIndex ];
 
 return true;
}

bool AdaBoost::recomputeNullRejectionThresholds(){
 
 if( trained ){
 //Todo - need to add null rejection for AdaBoost
 return false;
 }
 return false;
}

bool AdaBoost::setNullRejectionCoeff(Float nullRejectionCoeff){
 
 if( nullRejectionCoeff > 0 ){
 this->nullRejectionCoeff = nullRejectionCoeff;
 recomputeNullRejectionThresholds();
 return true;
 }
 return false;
}

bool AdaBoost::save( std::fstream &file ) const{
 
 if(!file.is_open())
 {
 errorLog <<"save(fstream &file) - The file is not open!" << std::endl;
 return false;
 }
 
 //Write the header info
 file<<"GRT_ADABOOST_MODEL_FILE_V2.0\n";
 
 //Write the classifier settings to the file
 if( !Classifier::saveBaseSettingsToFile(file) ){
 errorLog << __GRT_LOG__ << " Failed to save classifier base settings to file!" << std::endl;
 return false;
 }
 
 //Write the AdaBoost settings to the file
 file << "PredictionMethod: " << predictionMethod << std::endl;
 
 //If the model has been trained then write the model
 if( trained ){
 file << "Models: " << std::endl;
 for(UINT i=0; i<models.size(); i++){
 if( !models[i].save( file ) ){
 errorLog << __GRT_LOG__ << " Failed to write model " << i << " to file!" << std::endl;
 file.close();
 return false;
 }
 }
 }
 
 return true;
}

bool AdaBoost::load( std::fstream &file ){
 
 clear();
 
 if(!file.is_open())
 {
 errorLog << __GRT_LOG__ << " Could not open file to load model!" << std::endl;
 return false;
 }
 
 std::string word;
 file >> word;
 
 //Check to see if we should load a legacy file
 if( word == "GRT_ADABOOST_MODEL_FILE_V1.0" ){
 return loadLegacyModelFromFile( file );
 }
 
 if( word != "GRT_ADABOOST_MODEL_FILE_V2.0" ){
 errorLog << __GRT_LOG__ << " Failed to read file header!" << std::endl;
 errorLog << word << std::endl;
 return false;
 }
 
 //Load the base settings from the file
 if( !Classifier::loadBaseSettingsFromFile(file) ){
 errorLog << __GRT_LOG__ << " Failed to load base settings from file!" << std::endl;
 return false;
 }
 
 file >> word;
 if( word != "PredictionMethod:" ){
 errorLog << __GRT_LOG__ << " Failed to read PredictionMethod header!" << std::endl;
 return false;
 }
 file >> predictionMethod;
 
 if( trained ){
 file >> word;
 if( word != "Models:" ){
 errorLog << __GRT_LOG__ << " Failed to read Models header!" << std::endl;
 return false;
 }
 
 //Load the models
 models.resize( numClasses );
 for(UINT i=0; i<models.getSize(); i++){
 if( !models[i].load( file ) ){
 errorLog << __GRT_LOG__ << " Failed to load model " << i << " from file!" << std::endl;
 file.close();
 return false;
 }
 }
 
 //Recompute the null rejection thresholds
 recomputeNullRejectionThresholds();
 
 //Resize the prediction results to make sure it is setup for realtime prediction
 maxLikelihood = DEFAULT_NULL_LIKELIHOOD_VALUE;
 bestDistance = DEFAULT_NULL_DISTANCE_VALUE;
 classLikelihoods.resize(numClasses,DEFAULT_NULL_LIKELIHOOD_VALUE);
 classDistances.resize(numClasses,DEFAULT_NULL_DISTANCE_VALUE);
 }
 
 return true;
}

bool AdaBoost::clear(){
 
 //Clear the Classifier variables
 Classifier::clear();
 
 //Clear the AdaBoost model
 models.clear();
 
 return true;
}

bool AdaBoost::setWeakClassifier(const WeakClassifier &weakClassifer){
 
 //Clear any previous weak classifers
 clearWeakClassifiers();
 
 WeakClassifier *weakClassiferPtr = weakClassifer.createNewInstance();
 
 weakClassifiers.push_back( weakClassiferPtr );
 
 return true;
}

bool AdaBoost::addWeakClassifier(const WeakClassifier &weakClassifer){
 
 WeakClassifier *weakClassiferPtr = weakClassifer.createNewInstance();
 weakClassifiers.push_back( weakClassiferPtr );
 
 return true;
}

bool AdaBoost::clearWeakClassifiers(){
 
 for(UINT i=0; i<weakClassifiers.size(); i++){
 if( weakClassifiers[i] != NULL ){
 delete weakClassifiers[i];
 weakClassifiers[i] = NULL;
 }
 }
 weakClassifiers.clear();
 return true;
}

bool AdaBoost::setNumBoostingIterations(UINT numBoostingIterations){
 if( numBoostingIterations > 0 ){
 this->numBoostingIterations = numBoostingIterations;
 return true;
 }
 return false;
}

bool AdaBoost::setPredictionMethod(UINT predictionMethod){
 if( predictionMethod != MAX_POSITIVE_VALUE && predictionMethod != MAX_VALUE ){
 return false;
 }
 this->predictionMethod = predictionMethod;
 return true;
}

void AdaBoost::printModel(){
 
 std::cout <<"AdaBoostModel: \n";
 std::cout<<"NumFeatures: " << numInputDimensions << std::endl;
 std::cout<<"NumClasses: " << numClasses << std::endl;
 std::cout <<"UseScaling: " << useScaling << std::endl;
 std::cout<<"UseNullRejection: " << useNullRejection << std::endl;
 
 for(UINT k=0; k<numClasses; k++){
 std::cout << "Class: " << k+1 << " ClassLabel: " << classLabels[k] << std::endl;
 models[k].print();
 }
 
}

bool AdaBoost::loadLegacyModelFromFile( std::fstream &file ){
 
 std::string word;
 
 file >> word;
 if( word != "NumFeatures:" ){
 errorLog << __GRT_LOG__ << " Failed to read NumFeatures header!" << std::endl;
 return false;
 }
 file >> numInputDimensions;
 
 file >> word;
 if( word != "NumClasses:" ){
 errorLog << __GRT_LOG__ << " Failed to read NumClasses header!" << std::endl;
 return false;
 }
 file >> numClasses;
 
 file >> word;
 if( word != "UseScaling:" ){
 errorLog << __GRT_LOG__ << " Failed to read UseScaling header!" << std::endl;
 return false;
 }
 file >> useScaling;
 
 file >> word;
 if( word != "UseNullRejection:" ){
 errorLog << __GRT_LOG__ << " Failed to read UseNullRejection header!" << std::endl;
 return false;
 }
 file >> useNullRejection;
 
 if( useScaling ){
 file >> word;
 if( word != "Ranges:" ){
 errorLog << __GRT_LOG__ << " Failed to read Ranges header!" << std::endl;
 return false;
 }
 ranges.resize( numInputDimensions );
 
 for(UINT n=0; n<ranges.size(); n++){
 file >> ranges[n].minValue;
 file >> ranges[n].maxValue;
 }
 }
 
 file >> word;
 if( word != "Trained:" ){
 errorLog << __GRT_LOG__ << " Failed to read Trained header!" << std::endl;
 return false;
 }
 file >> trained;
 
 file >> word;
 if( word != "PredictionMethod:" ){
 errorLog << __GRT_LOG__ << " Failed to read PredictionMethod header!" << std::endl;
 return false;
 }
 file >> predictionMethod;
 
 if( trained ){
 file >> word;
 if( word != "Models:" ){
 errorLog << __GRT_LOG__ << " Failed to read Models header!" << std::endl;
 return false;
 }
 
 //Load the models
 models.resize( numClasses );
 classLabels.resize( numClasses );
 for(UINT i=0; i<models.getSize(); i++){
 if( !models[i].load( file ) ){
 errorLog << __GRT_LOG__ << " Failed to load model " << i << " from file!" << std::endl;
 file.close();
 return false;
 }
 
 //Set the class label
 classLabels[i] = models[i].getClassLabel();
 }
 }
 
 //Recompute the null rejection thresholds
 recomputeNullRejectionThresholds();
 
 //Resize the prediction results to make sure it is setup for realtime prediction
 maxLikelihood = DEFAULT_NULL_LIKELIHOOD_VALUE;
 bestDistance = DEFAULT_NULL_DISTANCE_VALUE;
 classLikelihoods.resize(numClasses,DEFAULT_NULL_LIKELIHOOD_VALUE);
 classDistances.resize(numClasses,DEFAULT_NULL_DISTANCE_VALUE);
 
 return true;
}

GRT_END_NAMESPACE