// Continue returns the results into ptRes for savePosteriors
// must be called the computeResult first!!!
void MDDAGClassifier::continueComputingResults(InputData* pData, vector<BaseLearner*>& weakHypotheses,
vector< ExampleResults* >& results, int fromIteration, int toIteration)
{
assert( !weakHypotheses.empty() );
const int numClasses = pData->getNumClasses();
const int numExamples = pData->getNumExamples();
// iterator over all the weak hypotheses
vector<BaseLearner*>::const_iterator whyIt;
int t;
for (whyIt = weakHypotheses.begin(), t = 0;
whyIt != weakHypotheses.end() && t < fromIteration; ++whyIt, ++t) {}
// for every feature: 1..T
for (; whyIt != weakHypotheses.end() && t < toIteration; ++whyIt, ++t)
{
BaseLearner* currWeakHyp = *whyIt;
AlphaReal alpha = currWeakHyp->getAlpha();
// for every point
for (int i = 0; i < numExamples; ++i)
{
// a reference for clarity and speed
vector<AlphaReal>& currVotesVector = results[i]->getVotesVector();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(pData, i, l);
}
}
}
void DataReader::calculateHypothesesMatrix()
{
cout << "[+] Calculate weak hyp matrix..." << endl;
const int numExamples = _pCurrentData->getNumExamples();
const int numClasses = _pCurrentData->getNumClasses();
hypermat& allOutputs = _weakHypothesesMatrices[_pCurrentData];
allOutputs.resize(numExamples);
cout << "Memory allocation for " << numExamples << " examples, " << _numIterations << " classifiers, and " << numClasses << " classes..." << flush;
for(int i = 0; i < numExamples; ++i)
{
allOutputs[i].resize(_numIterations);
for (int j = 0; j < _numIterations; ++j) {
allOutputs[i][j].resize(numClasses, 0.);
}
}
cout << "Done." << endl;
// const int step = (_totalNumIterations) < 50 ? 1 : (_totalNumIterations) / 50;
// cout << "Computing the weak hyp outputs: 0%." << flush;
cout << "Computing the weak hyp outputs... " << flush;
int t = 0;
for(int wHypInd = 0; wHypInd < _numIterations; ++wHypInd )
{
//
// if ((t + 1) % 1000 == 0)
// cout << "." << flush;
//
// if ((t + 1) % step == 0)
// {
// float progress = static_cast<float>(t) / (float)(_totalNumIterations) * 100.0;
// cout << "." << setprecision(2) << progress << "%." << flush;
// }
vector<BaseLearner*>::iterator whypIt;
for (whypIt = _weakHypotheses[wHypInd].begin(); whypIt != _weakHypotheses[wHypInd].end(); ++whypIt) {
// AbstainableLearner* currWeakHyp = dynamic_cast<AbstainableLearner*>(*whypIt);
BaseLearner* currWeakHyp = *whypIt;
AlphaReal alpha = currWeakHyp->getAlpha();
for(int i = 0; i < numExamples; ++i)
{
for (int l = 0; l < numClasses; ++l)
{
allOutputs[i][wHypInd][l] += alpha * currWeakHyp->classify(_pCurrentData, i, l);
}
}
++t;
}
}
cout << "Done." << endl;
}
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
bool AdaBoostMDPClassifier::classifyTestMDP( int i )
{
double acc=0.0;
const int numClasses = _pData->getNumClasses();
const int numExamples = _pTestData->getNumExamples();
ExampleResults* tmpResult = new ExampleResults( i, numClasses );
vector<AlphaReal>& currVotesVector = tmpResult->getVotesVector();
for( int j=0; j<_weakHypotheses.size(); j++ )
{
if (_history[j]) {
BaseLearner* currWeakHyp = _weakHypotheses[j];
float alpha = currWeakHyp->getAlpha();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(_pTestData, i, l);
}
}
vector<Label>::const_iterator lIt;
const vector<Label>& labels = _pTestData->getLabels(i);
// the vote of the winning negative class
float maxNegClass = -numeric_limits<float>::max();
// the vote of the winning positive class
float minPosClass = numeric_limits<float>::max();
for ( lIt = labels.begin(); lIt != labels.end(); ++lIt )
{
// get the negative winner class
if ( lIt->y < 0 && currVotesVector[lIt->idx] > maxNegClass )
maxNegClass = currVotesVector[lIt->idx];
// get the positive winner class
if ( lIt->y > 0 && currVotesVector[lIt->idx] < minPosClass )
minPosClass = currVotesVector[lIt->idx];
}
// if the vote for the worst positive label is lower than the
// vote for the highest negative label -> error
if (minPosClass <= maxNegClass)
return false;
else {
return true;
}
}
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
DataReader::DataReader(const nor_utils::Args& args, int verbose) : _verbose(verbose), _args(args)
{
string mdpTrainFileName = _args.getValue<string>("traintestmdp", 0);
string testFileName = _args.getValue<string>("traintestmdp", 1);
string shypFileName = _args.getValue<string>("traintestmdp", 3);
_numIterations = _args.getValue<int>("traintestmdp", 2);
string tmpFname = _args.getValue<string>("traintestmdp", 4);
if (_verbose > 0)
cout << "Loading arff data for MDP learning..." << flush;
// load the arff
loadInputData(mdpTrainFileName, testFileName, shypFileName);
if (_verbose > 0)
cout << "Done." << endl << flush;
if (_verbose > 0)
cout << "Loading strong hypothesis..." << flush;
// The class that loads the weak hypotheses
UnSerialization us;
// loads them
us.loadHypotheses(shypFileName, _weakHypotheses, _pTrainData);
if (_numIterations<_weakHypotheses.size())
_weakHypotheses.resize(_numIterations);
if (_verbose > 0)
cout << "Done." << endl << flush;
assert( _weakHypotheses.size() >= _numIterations );
// calculate the sum of alphas
vector<BaseLearner*>::iterator it;
_sumAlphas=0.0;
for( it = _weakHypotheses.begin(); it != _weakHypotheses.end(); ++it )
{
BaseLearner* currBLearner = *it;
_sumAlphas += currBLearner->getAlpha();
}
}
vector<AlphaReal> DataReader::getWhypClassification( const int wHypInd, const int instance )
{
const int numClasses = _pCurrentData->getNumClasses();
vector<AlphaReal> scoreVector(numClasses);
vector<BaseLearner*>::iterator whypIt;
for (whypIt = _weakHypotheses[wHypInd].begin(); whypIt != _weakHypotheses[wHypInd].end(); ++whypIt) {
BaseLearner* currWeakHyp = *whypIt;
AlphaReal alpha = currWeakHyp->getAlpha();
for (int l = 0; l < numClasses; ++l)
scoreVector[l] += alpha * currWeakHyp->classify(_pCurrentData, instance, l);
}
return scoreVector;
}
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
vector<int> DataReader::classifyKthWeakLearner( const int wHypInd, const int instance, ExampleResults* exampleResult )
{
if (_verbose>3) {
//cout << "Classifiying: " << wHypInd << endl;
}
if ( wHypInd >= _numIterations ) {
assert(false);
}
const int numClasses = _pCurrentData->getNumClasses();
// a reference for clarity and speed
vector<AlphaReal>& currVotesVector = exampleResult->getVotesVector();
vector<int> ternaryPhis(numClasses);
AlphaReal alpha;
// for every class
if (_isDataStorageMatrix)
{
for (int l = 0; l < numClasses; ++l) {
currVotesVector[l] += (*_pCurrentMatrix)[instance][wHypInd][l];
ternaryPhis[l] = (currVotesVector[l] > 0) ? 1 : ((currVotesVector[l] < 0) ? -1 : 0) ;
}
}
else
{
vector<BaseLearner*>::iterator whypIt;
for (whypIt = _weakHypotheses[wHypInd].begin(); whypIt != _weakHypotheses[wHypInd].end(); ++whypIt) {
BaseLearner* currWeakHyp = *whypIt;
alpha = currWeakHyp->getAlpha();
for (int l = 0; l < numClasses; ++l) {
int vote = currWeakHyp->classify(_pCurrentData, instance, l);
currVotesVector[l] += alpha * vote;
ternaryPhis[l] = (currVotesVector[l] > 0) ? 1 : ((currVotesVector[l] < 0) ? -1 : 0) ;
}
}
}
return ternaryPhis;
}
void SoftCascadeLearner::computePosteriors(InputData* pData, vector<BaseLearner*> & weakHypotheses, vector<AlphaReal> & oPosteriors, int positiveLabelIndex)
{
const int numExamples = pData->getNumExamples();
oPosteriors.resize(numExamples);
fill(oPosteriors.begin(), oPosteriors.end(), 0. );
vector<BaseLearner*>::iterator whyIt = weakHypotheses.begin();
for (;whyIt != weakHypotheses.end(); ++whyIt )
{
BaseLearner* currWeakHyp = *whyIt;
AlphaReal alpha = currWeakHyp->getAlpha();
for (int i = 0; i < numExamples; ++i)
{
AlphaReal alphaH = alpha * currWeakHyp->classify(pData, i, positiveLabelIndex);
oPosteriors[i] += alphaH;
}
}
}
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
double DataReader::classifyKthWeakLearner( const int wHypInd, const int instance, ExampleResults* exampleResult )
{
if (_verbose>3) {
//cout << "Classifiying: " << wHypInd << endl;
}
if ( wHypInd >= _numIterations ) return -1.0; // indicating error
const int numClasses = _pCurrentData->getNumClasses();
BaseLearner* currWeakHyp = _weakHypotheses[wHypInd];
float alpha = currWeakHyp->getAlpha();
// a reference for clarity and speed
vector<AlphaReal>& currVotesVector = exampleResult->getVotesVector();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(_pCurrentData, instance, l);
return alpha;
}
// Returns the results into ptRes
void MDDAGClassifier::computeResults(InputData* pData, vector<BaseLearner*>& weakHypotheses,
vector< ExampleResults* >& results, int numIterations)
{
assert( !weakHypotheses.empty() );
const int numClasses = pData->getNumClasses();
const int numExamples = pData->getNumExamples();
// Initialize the output info
OutputInfo* pOutInfo = NULL;
if ( !_outputInfoFile.empty() )
{
if ( _args.getNumValues("outputinfo") > 1 )
{
pOutInfo = new OutputInfo(_args);;
}
else
{
pOutInfo = new OutputInfo(_outputInfoFile, "e01hamauc", false);
}
}
// Creating the results structures. See file Structures.h for the
// PointResults structure
results.clear();
results.reserve(numExamples);
for (int i = 0; i < numExamples; ++i)
results.push_back( new ExampleResults(i, numClasses) );
// iterator over all the weak hypotheses
vector<BaseLearner*>::const_iterator whyIt;
int t;
if ( pOutInfo )
{
pOutInfo->initialize( pData );
pOutInfo->outputHeader(pData->getClassMap(),
true, // output iterations
false, // output time
true // endline
);
}
// for every feature: 1..T
for (whyIt = weakHypotheses.begin(), t = 0;
whyIt != weakHypotheses.end() && t < numIterations; ++whyIt, ++t)
{
BaseLearner* currWeakHyp = *whyIt;
AlphaReal alpha = currWeakHyp->getAlpha();
// for every point
for (int i = 0; i < numExamples; ++i)
{
// a reference for clarity and speed
vector<AlphaReal>& currVotesVector = results[i]->getVotesVector();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(pData, i, l);
}
// if needed output the step-by-step information
if ( pOutInfo )
{
pOutInfo->outputIteration(t);
// pOutInfo->outputError(pData, currWeakHyp);
// pOutInfo->outTPRFPR(pData);
//pOutInfo->outputBalancedError(pData, currWeakHyp);
// if ( ( t % 1 ) == 0 ) {
// pOutInfo->outputROC(pData);
// }
pOutInfo->outputCustom(pData, currWeakHyp);
// Margins and edge requires an update of the weight,
// therefore I keep them out for the moment
//outInfo.outputMargins(pData, currWeakHyp);
//outInfo.outputEdge(pData, currWeakHyp);
pOutInfo->endLine();
}
}
if (pOutInfo)
delete pOutInfo;
}
void MDDAGClassifier::saveLikelihoods(const string& dataFileName, const string& shypFileName,
const string& outFileName, int numIterations)
{
InputData* pData = loadInputData(dataFileName, shypFileName);
if (_verbose > 0)
cout << "Loading strong hypothesis..." << flush;
// The class that loads the weak hypotheses
UnSerialization us;
// Where to put the weak hypotheses
vector<BaseLearner*> weakHypotheses;
// loads them
us.loadHypotheses(shypFileName, weakHypotheses, pData);
// where the results go
vector< ExampleResults* > results;
if (_verbose > 0)
cout << "Classifying..." << flush;
const int numClasses = pData->getNumClasses();
const int numExamples = pData->getNumExamples();
ofstream outFile(outFileName.c_str());
string exampleName;
if (_verbose > 0)
cout << "Output likelihoods..." << flush;
// get the results
/////////////////////////////////////////////////////////////////////
// computeResults( pData, weakHypotheses, results, numIterations );
assert( !weakHypotheses.empty() );
// Initialize the output info
OutputInfo* pOutInfo = NULL;
if ( !_outputInfoFile.empty() )
pOutInfo = new OutputInfo(_outputInfoFile, "err");
// Creating the results structures. See file Structures.h for the
// PointResults structure
results.clear();
results.reserve(numExamples);
for (int i = 0; i < numExamples; ++i)
results.push_back( new ExampleResults(i, numClasses) );
// sum votes for classes
vector< AlphaReal > votesForExamples( numClasses );
vector< AlphaReal > expVotesForExamples( numClasses );
// iterator over all the weak hypotheses
vector<BaseLearner*>::const_iterator whyIt;
int t;
pOutInfo->initialize( pData );
// for every feature: 1..T
for (whyIt = weakHypotheses.begin(), t = 0;
whyIt != weakHypotheses.end() && t < numIterations; ++whyIt, ++t)
{
BaseLearner* currWeakHyp = *whyIt;
AlphaReal alpha = currWeakHyp->getAlpha();
// for every point
for (int i = 0; i < numExamples; ++i)
{
// a reference for clarity and speed
vector<AlphaReal>& currVotesVector = results[i]->getVotesVector();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(pData, i, l);
}
// if needed output the step-by-step information
if ( pOutInfo )
{
pOutInfo->outputIteration(t);
pOutInfo->outputCustom(pData, currWeakHyp);
// Margins and edge requires an update of the weight,
// therefore I keep them out for the moment
//outInfo.outputMargins(pData, currWeakHyp);
//outInfo.outputEdge(pData, currWeakHyp);
pOutInfo->endLine();
} // for (int i = 0; i < numExamples; ++i)
// calculate likelihoods from votes
fill( votesForExamples.begin(), votesForExamples.end(), 0.0 );
AlphaReal lLambda = 0.0;
for (int i = 0; i < numExamples; ++i)
{
// a reference for clarity and speed
vector<AlphaReal>& currVotesVector = results[i]->getVotesVector();
AlphaReal sumExp = 0.0;
// for every class
for (int l = 0; l < numClasses; ++l)
{
expVotesForExamples[l] = exp( currVotesVector[l] ) ;
sumExp += expVotesForExamples[l];
}
if ( sumExp > numeric_limits<AlphaReal>::epsilon() )
{
for (int l = 0; l < numClasses; ++l)
{
expVotesForExamples[l] /= sumExp;
}
}
Example ex = pData->getExample( results[i]->getIdx() );
vector<Label> labs = ex.getLabels();
AlphaReal m = numeric_limits<AlphaReal>::infinity();
for (int l = 0; l < numClasses; ++l)
{
if ( labs[l].y > 0 )
{
if ( expVotesForExamples[l] > numeric_limits<AlphaReal>::epsilon() )
{
AlphaReal logVal = log( expVotesForExamples[l] );
if ( logVal != m ) {
lLambda += ( ( 1.0/(AlphaReal)numExamples ) * logVal );
}
}
}
}
}
outFile << t << "\t" << lLambda ;
outFile << '\n';
outFile.flush();
}
if (pOutInfo)
delete pOutInfo;
// computeResults( pData, weakHypotheses, results, numIterations );
///////////////////////////////////////////////////////////////////////////////////
/*
for (int i = 0; i < numExamples; ++i)
{
// output the name if it exists, otherwise the number
// of the example
exampleName = pData->getExampleName(i);
if ( !exampleName.empty() )
outFile << exampleName << ',';
// output the posteriors
outFile << results[i]->getVotesVector()[0];
for (int l = 1; l < numClasses; ++l)
outFile << ',' << results[i]->getVotesVector()[l];
outFile << '\n';
}
*/
if (_verbose > 0)
cout << "Done!" << endl;
if (_verbose > 1)
{
cout << "\nClass order (You can change it in the header of the data file):" << endl;
for (int l = 0; l < numClasses; ++l)
cout << "- " << pData->getClassMap().getNameFromIdx(l) << endl;
}
// delete the input data file
if (pData)
delete pData;
vector<ExampleResults*>::iterator it;
for (it = results.begin(); it != results.end(); ++it)
delete (*it);
}
void FilterBoostLearner::run(const nor_utils::Args& args)
{
// load the arguments
this->getArgs(args);
time_t startTime, currentTime;
time(&startTime);
// get the registered weak learner (type from name)
BaseLearner* pWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner(_baseLearnerName);
// initialize learning options; normally it's done in the strong loop
// also, here we do it for Product learners, so input data can be created
pWeakHypothesisSource->initLearningOptions(args);
BaseLearner* pConstantWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner("ConstantLearner");
// get the training input data, and load it
InputData* pTrainingData = pWeakHypothesisSource->createInputData();
pTrainingData->initOptions(args);
pTrainingData->load(_trainFileName, IT_TRAIN, _verbose);
const int numClasses = pTrainingData->getNumClasses();
const int numExamples = pTrainingData->getNumExamples();
//initialize the margins variable
_margins.resize( numExamples );
for( int i=0; i<numExamples; i++ )
{
_margins[i].resize( numClasses );
fill( _margins[i].begin(), _margins[i].end(), 0.0 );
}
// get the testing input data, and load it
InputData* pTestData = NULL;
if ( !_testFileName.empty() )
{
pTestData = pWeakHypothesisSource->createInputData();
pTestData->initOptions(args);
pTestData->load(_testFileName, IT_TEST, _verbose);
}
// The output information object
OutputInfo* pOutInfo = NULL;
if ( !_outputInfoFile.empty() )
{
// Baseline: constant classifier - goes into 0th iteration
BaseLearner* pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal constantEnergy = pConstantWeakHypothesis->run();
pOutInfo = new OutputInfo(args);
pOutInfo->initialize(pTrainingData);
updateMargins( pTrainingData, pConstantWeakHypothesis );
if (pTestData)
pOutInfo->initialize(pTestData);
pOutInfo->outputHeader(pTrainingData->getClassMap() );
pOutInfo->outputIteration(-1);
pOutInfo->outputCustom(pTrainingData, pConstantWeakHypothesis);
if (pTestData)
{
pOutInfo->separator();
pOutInfo->outputCustom(pTestData, pConstantWeakHypothesis);
}
pOutInfo->outputCurrentTime();
pOutInfo->endLine();
pOutInfo->initialize(pTrainingData);
if (pTestData)
pOutInfo->initialize(pTestData);
}
// reload the previously found weak learners if -resume is set.
// otherwise just return 0
int startingIteration = resumeWeakLearners(pTrainingData);
Serialization ss(_shypFileName, _isShypCompressed );
ss.writeHeader(_baseLearnerName); // this must go after resumeProcess has been called
// perform the resuming if necessary. If not it will just return
resumeProcess(ss, pTrainingData, pTestData, pOutInfo);
if (_verbose == 1)
cout << "Learning in progress..." << endl;
///////////////////////////////////////////////////////////////////////
// Starting the AdaBoost main loop
///////////////////////////////////////////////////////////////////////
for (int t = startingIteration; t < _numIterations; ++t)
{
if (_verbose > 1)
cout << "------- WORKING ON ITERATION " << (t+1) << " -------" << endl;
// create the weak learner
BaseLearner* pWeakHypothesis;
BaseLearner* pConstantWeakHypothesis;
pWeakHypothesis = pWeakHypothesisSource->create();
pWeakHypothesis->initLearningOptions(args);
//pTrainingData->clearIndexSet();
pWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal edge, energy=0.0;
// create the constant learner
pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal constantEdge = -numeric_limits<AlphaReal>::max();
int currentNumberOfUsedData = static_cast<int>(_Cn * log(t+3.0));
if ( _onlineWeakLearning )
{
//check whether the weak learner is a ScalarLeaerner
try {
StochasticLearner* pStochasticLearner = dynamic_cast<StochasticLearner*>(pWeakHypothesis);
StochasticLearner* pStochasticConstantWeakHypothesis = dynamic_cast<StochasticLearner*> (pConstantWeakHypothesis);
pStochasticLearner->initLearning();
pStochasticConstantWeakHypothesis->initLearning();
if (_verbose>1)
cout << "Number of random instances: \t" << currentNumberOfUsedData << endl;
// set the weights
setWeightToMargins(pTrainingData);
//learning
for (int i=0; i<currentNumberOfUsedData; ++i )
{
int randomIndex = (rand() % pTrainingData->getNumExamples());
//int randomIndex = getRandomIndex();
pStochasticLearner->update(randomIndex);
pStochasticConstantWeakHypothesis->update(randomIndex);
}
pStochasticLearner->finishLearning();
pStochasticConstantWeakHypothesis->finishLearning();
}
catch (bad_cast& e) {
cerr << "The weak learner must be a StochasticLearner!!!" << endl;
exit(-1);
}
}
else
{
filter( pTrainingData, currentNumberOfUsedData );
if ( pTrainingData->getNumExamples() < 2 )
{
filter( pTrainingData, currentNumberOfUsedData, false );
}
if (_verbose > 1)
{
cout << "--> Size of training data = " << pTrainingData->getNumExamples() << endl;
}
energy = pWeakHypothesis->run();
pConstantWeakHypothesis->run();
}
//estimate edge
filter( pTrainingData, currentNumberOfUsedData, false );
edge = pWeakHypothesis->getEdge(true) / 2.0;
constantEdge = pConstantWeakHypothesis->getEdge() / 2.0;
if ( constantEdge > edge )
{
delete pWeakHypothesis;
pWeakHypothesis = pConstantWeakHypothesis;
edge = constantEdge;
} else {
delete pConstantWeakHypothesis;
}
// calculate alpha
AlphaReal alpha = 0.0;
alpha = 0.5 * log( ( 1 + edge ) / ( 1 - edge ) );
pWeakHypothesis->setAlpha( alpha );
_sumAlpha += alpha;
if (_verbose > 1)
cout << "Weak learner: " << pWeakHypothesis->getName()<< endl;
// Output the step-by-step information
pTrainingData->clearIndexSet();
printOutputInfo(pOutInfo, t, pTrainingData, pTestData, pWeakHypothesis);
// Updates the weights and returns the edge
//AlphaReal gamma = updateWeights(pTrainingData, pWeakHypothesis);
if (_verbose > 1)
{
cout << setprecision(5)
<< "--> Alpha = " << pWeakHypothesis->getAlpha() << endl
<< "--> Edge = " << edge << endl
<< "--> Energy = " << energy << endl
// << "--> ConstantEnergy = " << constantEnergy << endl
// << "--> difference = " << (energy - constantEnergy) << endl
;
}
// update the margins
//saveMargins();
updateMargins( pTrainingData, pWeakHypothesis );
// append the current weak learner to strong hypothesis file,
// that is, serialize it.
ss.appendHypothesis(t, pWeakHypothesis);
// Add it to the internal list of weak hypotheses
_foundHypotheses.push_back(pWeakHypothesis);
// check if the time limit has been reached
if (_maxTime > 0)
{
time( ¤tTime );
float diff = difftime(currentTime, startTime); // difftime is in seconds
diff /= 60; // = minutes
if (diff > _maxTime)
{
if (_verbose > 0)
cout << "Time limit of " << _maxTime
<< " minutes has been reached!" << endl;
break;
}
} // check for maxtime
delete pWeakHypothesis;
} // loop on iterations
/////////////////////////////////////////////////////////
// write the footer of the strong hypothesis file
ss.writeFooter();
// Free the two input data objects
if (pTrainingData)
delete pTrainingData;
if (pTestData)
delete pTestData;
if (pOutInfo)
delete pOutInfo;
if (_verbose > 0)
cout << "Learning completed." << endl;
}
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
double DataReader::getAdaboostPerfOnCurrentDataset()
{
const int numClasses = _pCurrentData->getNumClasses();
const int numExamples = _pCurrentData->getNumExamples();
int correct = 0;
int incorrect = 0;
double err;
vector<double>& iterationWiseError = _iterationWiseError[_pCurrentData];
iterationWiseError.resize(_weakHypotheses.size(), 0.);
vector<ExampleResults*> examplesResults(numExamples);
for (int i = 0; i < numExamples; ++i)
examplesResults[i] = new ExampleResults(i, numClasses) ;
for( int j = 0; j < _weakHypotheses.size(); ++j )
{
correct = 0;
incorrect = 0;
for( int i = 0; i < numExamples; ++i )
{
ExampleResults*& tmpResult = examplesResults[i];
vector<AlphaReal>& currVotesVector = tmpResult->getVotesVector();
if (_isDataStorageMatrix)
{
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += (*_pCurrentMatrix)[i][j][l];
}
else
{
vector<BaseLearner*>::iterator whypIt;
for (whypIt = _weakHypotheses[j].begin(); whypIt != _weakHypotheses[j].end(); ++whypIt) {
BaseLearner* currWeakHyp = *whypIt;
AlphaReal alpha = currWeakHyp->getAlpha();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(_pCurrentData, i, l);
}
}
vector<Label>::const_iterator lIt;
const vector<Label>& labels = _pCurrentData->getLabels(i);
// the vote of the winning negative class
AlphaReal maxNegClass = -numeric_limits<AlphaReal>::max();
// the vote of the winning positive class
AlphaReal minPosClass = numeric_limits<AlphaReal>::max();
for ( lIt = labels.begin(); lIt != labels.end(); ++lIt )
{
// get the negative winner class
if ( lIt->y < 0 && currVotesVector[lIt->idx] > maxNegClass )
maxNegClass = currVotesVector[lIt->idx];
// get the positive winner class
if ( lIt->y > 0 && currVotesVector[lIt->idx] < minPosClass )
minPosClass = currVotesVector[lIt->idx];
}
// if the vote for the worst positive label is lower than the
// vote for the highest negative label -> error
if (minPosClass <= maxNegClass)
incorrect++;
else {
correct++;
}
}
err = ((double) incorrect / ((double) numExamples)); // * 100.0;
iterationWiseError[j] = err;
}
// cout << endl;
// int i = 0;
// for (const auto & myTmpKey : _iterationWiseError[_pCurrentData]) {
// cout << myTmpKey << " ";
// ++i;
// if (i > 50) {
// break;
// }
// }
// cout << endl;
for (int i = 0; i < numExamples; ++i)
delete examplesResults[i] ;
// double acc = ((double) correct / ((double) numExamples)) * 100.0;
return err;
}
void SoftCascadeLearner::run(const nor_utils::Args& args)
{
// load the arguments
this->getArgs(args);
//print cascade properties
if (_verbose > 0) {
cout << "[+] Softcascade parameters :" << endl
<< "\t --> target detection rate = " << _targetDetectionRate << endl
<< "\t --> alpha (exp param) = " << _alphaExponentialParameter << endl
<< "\t --> bootstrap rate = " << _bootstrapRate << endl
<< endl;
}
// get the registered weak learner (type from name)
BaseLearner* pWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner(_baseLearnerName);
// initialize learning options; normally it's done in the strong loop
// also, here we do it for Product learners, so input data can be created
pWeakHypothesisSource->initLearningOptions(args);
// get the training input data, and load it
InputData* pTrainingData = pWeakHypothesisSource->createInputData();
pTrainingData->initOptions(args);
pTrainingData->load(_trainFileName, IT_TRAIN, 5);
InputData* pBootstrapData = NULL;
if (!_bootstrapFileName.empty()) {
pBootstrapData = pWeakHypothesisSource->createInputData();
pBootstrapData->initOptions(args);
pBootstrapData->load(_bootstrapFileName, IT_TRAIN, 5);
}
// get the testing input data, and load it
InputData* pTestData = NULL;
if ( !_testFileName.empty() )
{
pTestData = pWeakHypothesisSource->createInputData();
pTestData->initOptions(args);
pTestData->load(_testFileName, IT_TEST, 5);
}
Serialization ss(_shypFileName, false );
ss.writeHeader(_baseLearnerName);
// outputHeader();
// The output information object
OutputInfo* pOutInfo = NULL;
if ( !_outputInfoFile.empty() )
{
pOutInfo = new OutputInfo(args, true);
pOutInfo->setOutputList("sca", &args);
pOutInfo->initialize(pTrainingData);
if (pTestData)
pOutInfo->initialize(pTestData);
pOutInfo->outputHeader(pTrainingData->getClassMap(), true, true, false);
pOutInfo->outputUserHeader("thresh");
pOutInfo->headerEndLine();
}
// ofstream trainPosteriorsFile;
// ofstream testPosteriorsFile;
const NameMap& namemap = pTrainingData->getClassMap();
_positiveLabelIndex = namemap.getIdxFromName(_positiveLabelName);
// FIXME: output posteriors
// OutputInfo* pTrainPosteriorsOut = NULL;
// OutputInfo* pTestPosteriorsOut = NULL;
// if (! _trainPosteriorsFileName.empty()) {
// pTrainPosteriorsOut = new OutputInfo(_trainPosteriorsFileName, "pos", true);
// pTrainPosteriorsOut->initialize(pTrainingData);
// dynamic_cast<PosteriorsOutput*>( pTrainPosteriorsOut->getOutputInfoObject("pos") )->addClassIndex(_positiveLabelIndex );
// }
// if (! _testPosteriorsFileName.empty() && !_testFileName.empty() ) {
// pTestPosteriorsOut = new OutputInfo(_testPosteriorsFileName, "pos", true);
// pTestPosteriorsOut->initialize(pTestData);
// dynamic_cast<PosteriorsOutput*>( pTestPosteriorsOut->getOutputInfoObject("pos") )->addClassIndex(_positiveLabelIndex );
// }
const int numExamples = pTrainingData->getNumExamples();
vector<BaseLearner*> inWeakHypotheses;
if (_fullRun) {
// TODO : the full training is implementet, testing is needed
AdaBoostMHLearner* sHypothesis = new AdaBoostMHLearner();
sHypothesis->run(args, pTrainingData, _baseLearnerName, _numIterations, inWeakHypotheses );
delete sHypothesis;
}
else {
cout << "[+] Loading uncalibrated shyp file... ";
//read the shyp file of the trained classifier
UnSerialization us;
us.loadHypotheses(_unCalibratedShypFileName, inWeakHypotheses, pTrainingData);
if (_inShypLimit > 0 && _inShypLimit < inWeakHypotheses.size() ) {
inWeakHypotheses.resize(_inShypLimit);
}
if (_numIterations > inWeakHypotheses.size()) {
_numIterations = inWeakHypotheses.size();
}
cout << "weak hypotheses loaded, " << inWeakHypotheses.size() << " retained.\n";
}
// some initializations
_foundHypotheses.resize(0);
double faceRejectionFraction = 0.;
double estimatedExecutionTime = 0.;
vector<double> rejectionDistributionVector;
_rejectionThresholds.resize(0);
set<int> trainingIndices;
for (int i = 0; i < numExamples; i++) {
trainingIndices.insert(pTrainingData->getRawIndex(i) );
}
// init v_t (see the paper)
initializeRejectionDistributionVector(_numIterations, rejectionDistributionVector);
if (_verbose == 1)
cout << "Learning in progress..." << endl;
///////////////////////////////////////////////////////////////////////
// Starting the SoftCascade main loop
///////////////////////////////////////////////////////////////////////
for (int t = 0; t < _numIterations; ++t)
{
if (_verbose > 0)
cout << "--------------[ iteration " << (t+1) << " ]--------------" << endl;
faceRejectionFraction += rejectionDistributionVector[t];
cout << "[+] Face rejection tolerated : " << faceRejectionFraction << " | v[t] = " << rejectionDistributionVector[t] << endl;
int numberOfNegatives = pTrainingData->getNumExamplesPerClass(1 - _positiveLabelIndex);
//vector<BaseLearner*>::const_iterator whyIt;
int selectedIndex = 0;
AlphaReal bestGap = 0;
vector<AlphaReal> posteriors;
computePosteriors(pTrainingData, _foundHypotheses, posteriors, _positiveLabelIndex);
//should use an iterator instead of i
vector<BaseLearner*>::iterator whyIt;
int i;
for (i = 0, whyIt = inWeakHypotheses.begin(); whyIt != inWeakHypotheses.end(); ++whyIt, ++i) {
vector<AlphaReal> temporaryPosteriors = posteriors;
vector<BaseLearner*> temporaryWeakHyp = _foundHypotheses;
temporaryWeakHyp.push_back(*whyIt);
updatePosteriors(pTrainingData, *whyIt, temporaryPosteriors, _positiveLabelIndex);
AlphaReal gap = computeSeparationSpan(pTrainingData, temporaryPosteriors, _positiveLabelIndex );
if (gap > bestGap) {
bestGap = gap;
selectedIndex = i;
}
}
BaseLearner* selectedWeakHypothesis = inWeakHypotheses[selectedIndex];
cout << "[+] Rank of the selected weak hypothesis : " << selectedIndex << endl
<< "\t ---> edge gap = " << bestGap << endl
<< "\t ---> alpha = " << selectedWeakHypothesis->getAlpha() << endl;
//update the stages
_foundHypotheses.push_back(selectedWeakHypothesis);
updatePosteriors(pTrainingData, selectedWeakHypothesis, posteriors, _positiveLabelIndex);
double missesFraction;
AlphaReal r = findBestRejectionThreshold(pTrainingData, posteriors, faceRejectionFraction, missesFraction);
_rejectionThresholds.push_back(r);
// update the output info object
dynamic_cast<SoftCascadeOutput*>( pOutInfo->getOutputInfoObject("sca") )->appendRejectionThreshold(r);
cout << "[+] Rejection threshold = " << r << endl;
//some updates
ss.appendHypothesisWithThreshold(t, selectedWeakHypothesis, r);
faceRejectionFraction -= missesFraction;
inWeakHypotheses.erase(inWeakHypotheses.begin() + selectedIndex);
double whypCost = 1; //just in case there are different costs for each whyp
estimatedExecutionTime += whypCost * numberOfNegatives;
// output perf in file
vector< vector< AlphaReal> > scores(0);
_output << t + 1 << setw(_sepWidth + 1) << r << setw(_sepWidth);
// update OutputInfo with the new whyp
// updateOutputInfo(pOutInfo, pTrainingData, selectedWeakHypothesis);
// if (pTestData) {
// updateOutputInfo(pOutInfo, pTestData, selectedWeakHypothesis);
// }
// output the iteration results
printOutputInfo(pOutInfo, t, pTrainingData, pTestData, selectedWeakHypothesis, r);
// if (pTrainPosteriorsOut) {
// pTrainPosteriorsOut->setTable(pTrainingData, pOutInfo->getTable(pTrainingData));
// pTrainPosteriorsOut->outputCustom(pTrainingData);
// }
//
// if (pTestPosteriorsOut) {
// pTestPosteriorsOut->setTable(pTestData, pOutInfo->getTable(pTestData));
// pTestPosteriorsOut->outputCustom(pTestData);
// }
int leftNegatives = filterDataset(pTrainingData, posteriors, r, trainingIndices);
if (leftNegatives == 0) {
cout << endl << "[+] No more negatives.\n";
break;
}
if (_bootstrapRate != 0) {
bootstrapTrainingSet(pTrainingData, pBootstrapData, trainingIndices);
}
} // loop on iterations
/////////////////////////////////////////////////////////
// write the footer of the strong hypothesis file
ss.writeFooter();
// Free the two input data objects
if (pTrainingData)
delete pTrainingData;
if (pBootstrapData) {
delete pBootstrapData;
}
if (pTestData)
delete pTestData;
if (_verbose > 0)
cout << "Learning completed." << endl;
}
void FilterBoostLearner::run(const nor_utils::Args& args)
{
// load the arguments
this->getArgs(args);
time_t startTime, currentTime;
time(&startTime);
// get the registered weak learner (type from name)
BaseLearner* pWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner(_baseLearnerName);
// initialize learning options; normally it's done in the strong loop
// also, here we do it for Product learners, so input data can be created
pWeakHypothesisSource->initLearningOptions(args);
BaseLearner* pConstantWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner("ConstantLearner");
// get the training input data, and load it
InputData* pTrainingData = pWeakHypothesisSource->createInputData();
pTrainingData->initOptions(args);
pTrainingData->load(_trainFileName, IT_TRAIN, _verbose);
const int numClasses = pTrainingData->getNumClasses();
const int numExamples = pTrainingData->getNumExamples();
//initialize the margins variable
_margins.resize( numExamples );
for( int i=0; i<numExamples; i++ )
{
_margins[i].resize( numClasses );
fill( _margins[i].begin(), _margins[i].end(), 0.0 );
}
// get the testing input data, and load it
InputData* pTestData = NULL;
if ( !_testFileName.empty() )
{
pTestData = pWeakHypothesisSource->createInputData();
pTestData->initOptions(args);
pTestData->load(_testFileName, IT_TEST, _verbose);
}
// The output information object
OutputInfo* pOutInfo = NULL;
if ( !_outputInfoFile.empty() )
{
// Baseline: constant classifier - goes into 0th iteration
BaseLearner* pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
float constantEnergy = pConstantWeakHypothesis->run();
pOutInfo = new OutputInfo(_outputInfoFile);
pOutInfo->initialize(pTrainingData);
updateMargins( pTrainingData, pConstantWeakHypothesis );
if (pTestData)
pOutInfo->initialize(pTestData);
pOutInfo->outputHeader();
pOutInfo->outputIteration(-1);
pOutInfo->outputError(pTrainingData, pConstantWeakHypothesis);
if (pTestData)
pOutInfo->outputError(pTestData, pConstantWeakHypothesis);
/*
pOutInfo->outputMargins(pTrainingData, pConstantWeakHypothesis);
pOutInfo->outputEdge(pTrainingData, pConstantWeakHypothesis);
if (pTestData)
pOutInfo->outputMargins(pTestData, pConstantWeakHypothesis);
pOutInfo->outputMAE(pTrainingData);
if (pTestData)
pOutInfo->outputMAE(pTestData);
*/
pOutInfo->outputCurrentTime();
pOutInfo->endLine();
pOutInfo->initialize(pTrainingData);
if (pTestData)
pOutInfo->initialize(pTestData);
}
// reload the previously found weak learners if -resume is set.
// otherwise just return 0
int startingIteration = resumeWeakLearners(pTrainingData);
Serialization ss(_shypFileName, _isShypCompressed );
ss.writeHeader(_baseLearnerName); // this must go after resumeProcess has been called
// perform the resuming if necessary. If not it will just return
resumeProcess(ss, pTrainingData, pTestData, pOutInfo);
if (_verbose == 1)
cout << "Learning in progress..." << endl;
///////////////////////////////////////////////////////////////////////
// Starting the AdaBoost main loop
///////////////////////////////////////////////////////////////////////
for (int t = startingIteration; t < _numIterations; ++t)
{
if (_verbose > 1)
cout << "------- WORKING ON ITERATION " << (t+1) << " -------" << endl;
filter( pTrainingData, (int)(_Cn * log(t+2.0)) );
if ( pTrainingData->getNumExamples() < 2 )
{
filter( pTrainingData, (int)(_Cn * log(t+2.0)), false );
}
if (_verbose > 1)
{
cout << "--> Size of training data = " << pTrainingData->getNumExamples() << endl;
}
BaseLearner* pWeakHypothesis = pWeakHypothesisSource->create();
pWeakHypothesis->initLearningOptions(args);
//pTrainingData->clearIndexSet();
pWeakHypothesis->setTrainingData(pTrainingData);
float energy = pWeakHypothesis->run();
BaseLearner* pConstantWeakHypothesis;
if (_withConstantLearner) // check constant learner if user wants it
{
pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
float constantEnergy = pConstantWeakHypothesis->run();
}
//estimate edge
filter( pTrainingData, (int)(_Cn * log(t+2.0)), false );
float edge = pWeakHypothesis->getEdge() / 2.0;
if (_withConstantLearner) // check constant learner if user wants it
{
float constantEdge = pConstantWeakHypothesis->getEdge() / 2.0;
if ( constantEdge > edge )
{
delete pWeakHypothesis;
pWeakHypothesis = pConstantWeakHypothesis;
edge = constantEdge;
} else {
delete pConstantWeakHypothesis;
}
}
// calculate alpha
float alpha = 0.0;
alpha = 0.5 * log( ( 0.5 + edge ) / ( 0.5 - edge ) );
pWeakHypothesis->setAlpha( alpha );
if (_verbose > 1)
cout << "Weak learner: " << pWeakHypothesis->getName()<< endl;
// Output the step-by-step information
pTrainingData->clearIndexSet();
printOutputInfo(pOutInfo, t, pTrainingData, pTestData, pWeakHypothesis);
// Updates the weights and returns the edge
float gamma = updateWeights(pTrainingData, pWeakHypothesis);
if (_verbose > 1)
{
cout << setprecision(5)
<< "--> Alpha = " << pWeakHypothesis->getAlpha() << endl
<< "--> Edge = " << gamma << endl
<< "--> Energy = " << energy << endl
// << "--> ConstantEnergy = " << constantEnergy << endl
// << "--> difference = " << (energy - constantEnergy) << endl
;
}
// update the margins
updateMargins( pTrainingData, pWeakHypothesis );
// append the current weak learner to strong hypothesis file,
// that is, serialize it.
ss.appendHypothesis(t, pWeakHypothesis);
// Add it to the internal list of weak hypotheses
_foundHypotheses.push_back(pWeakHypothesis);
// check if the time limit has been reached
if (_maxTime > 0)
{
time( ¤tTime );
float diff = difftime(currentTime, startTime); // difftime is in seconds
diff /= 60; // = minutes
if (diff > _maxTime)
{
if (_verbose > 0)
cout << "Time limit of " << _maxTime
<< " minutes has been reached!" << endl;
break;
}
} // check for maxtime
delete pWeakHypothesis;
} // loop on iterations
/////////////////////////////////////////////////////////
// write the footer of the strong hypothesis file
ss.writeFooter();
// Free the two input data objects
if (pTrainingData)
delete pTrainingData;
if (pTestData)
delete pTestData;
if (pOutInfo)
delete pOutInfo;
if (_verbose > 0)
cout << "Learning completed." << endl;
}
// -------------------------------------------------------------------------
void AdaBoostMHLearner::run( const nor_utils::Args& args, InputData* pTrainingData, const string baseLearnerName, const int numIterations, vector<BaseLearner*>& foundHypotheses )
{
// get the registered weak learner (type from name)
BaseLearner* pWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner(baseLearnerName);
// initialize learning options; normally it's done in the strong loop
// also, here we do it for Product learners, so input data can be created
pWeakHypothesisSource->initLearningOptions(args);
BaseLearner* pConstantWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner("ConstantLearner");
if (_verbose == 1)
cout << "Learning in progress..." << endl;
///////////////////////////////////////////////////////////////////////
// Starting the AdaBoost main loop
///////////////////////////////////////////////////////////////////////
for (int t = 0; t < numIterations; ++t)
{
if ((_verbose > 0)&&((t%100)==0))
cout << "--------------[ Boosting iteration " << (t+1) << " ]--------------" << endl;
BaseLearner* pWeakHypothesis = pWeakHypothesisSource->create();
pWeakHypothesis->initLearningOptions(args);
//pTrainingData->clearIndexSet();
pWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal energy = pWeakHypothesis->run();
//float gamma = pWeakHypothesis->getEdge();
//cout << gamma << endl;
if ( (_withConstantLearner) || ( energy != energy ) ) // check constant learner if user wants it (if energi is nan, then we chose constant learner
{
BaseLearner* pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal constantEnergy = pConstantWeakHypothesis->run();
if ( (constantEnergy <= energy) || ( energy != energy ) ) {
delete pWeakHypothesis;
pWeakHypothesis = pConstantWeakHypothesis;
}
}
if (_verbose > 1)
cout << "Weak learner: " << pWeakHypothesis->getName()<< endl;
// Updates the weights and returns the edge
AlphaReal gamma = updateWeights(pTrainingData, pWeakHypothesis);
if (_verbose > 1)
{
cout << setprecision(5)
<< "--> Alpha = " << pWeakHypothesis->getAlpha() << endl
<< "--> Edge = " << gamma << endl
<< "--> Energy = " << energy << endl
// << "--> ConstantEnergy = " << constantEnergy << endl
// << "--> difference = " << (energy - constantEnergy) << endl
;
}
// If gamma <= theta the algorithm must stop.
// If theta == 0 and gamma is 0, it means that the weak learner is no better than chance
// and no further training is possible.
if (gamma <= _theta)
{
if (_verbose > 0)
{
cout << "Can't train any further: edge = " << gamma
<< " (with and edge offset (theta)=" << _theta << ")" << endl;
}
// delete pWeakHypothesis;
// break;
}
// Add it to the internal list of weak hypotheses
foundHypotheses.push_back(pWeakHypothesis);
} // loop on iterations
/////////////////////////////////////////////////////////
if (_verbose > 0)
cout << "--------------[ AdaBoost Learning completed. ]--------------" << endl;
}
// -------------------------------------------------------------------------
// -------------------------------------------------------------------------
AlphaReal AdaBoostMHLearner::updateWeights(OutputInfo* pOutInfo, InputData* pData, vector<BaseLearner*>& pWeakHypothesis){
const int numExamples = pData->getNumExamples();
const int numClasses = pData->getNumClasses();
AlphaReal Z = 0; // The normalization factor
// _hy will contain the margins
_hy.resize(numExamples);
for ( int i = 0; i < numExamples; ++i){
_hy[i].resize(numClasses);
fill( _hy[i].begin(), _hy[i].end(), 0.0 );
}
vector<BaseLearner*>::iterator it;
if (_verbose > 0)
cout << ": 0%." << flush;
const int numIters = static_cast<int>(_foundHypotheses.size());
const int step = numIters < 5 ? 1 : numIters / 5;
int t = 0;
// calculate the margins ( f^{t}(x_i) ), _hy will contain
for( it = pWeakHypothesis.begin(); it != pWeakHypothesis.end(); it++, t++ )
{
if (_verbose > 1 && (t + 1) % step == 0)
{
float progress = static_cast<float>(t) / static_cast<float>(numIters) * 100.0;
cout << "." << setprecision(2) << progress << "%." << flush;
}
BaseLearner* pWeakHypothesis = *it;
const AlphaReal alpha = pWeakHypothesis->getAlpha();
AlphaReal hx;
for (int i = 0; i < numExamples; ++i)
{
vector<Label>& labels = pData->getLabels(i);
vector<Label>::iterator lIt;
for (lIt = labels.begin(); lIt != labels.end(); ++lIt )
{
hx = pWeakHypothesis->classify(pData, i, lIt->idx );
_hy[i][lIt->idx] += alpha * hx; // alpha * h_l(x_i)
lIt->weight *= exp( -alpha * hx * lIt->y );
Z += lIt->weight;
}
}
// renormalize the weights
for (int i = 0; i < numExamples; ++i)
{
vector<Label>& labels = pData->getLabels(i);
vector<Label>::iterator lIt;
for (lIt = labels.begin(); lIt != labels.end(); ++lIt )
{
lIt->weight /= Z;
}
}
//i++;
//if ( i % 1000 == 0 ) cout << i <<endl;
}
//upload the margins
pOutInfo->setTable( pData, _hy );
return 0;
}
void AdaBoostMHLearner::run(const nor_utils::Args& args)
{
// load the arguments
this->getArgs(args);
// get the registered weak learner (type from name)
BaseLearner* pWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner(_baseLearnerName);
// initialize learning options; normally it's done in the strong loop
// also, here we do it for Product learners, so input data can be created
pWeakHypothesisSource->initLearningOptions(args);
BaseLearner* pConstantWeakHypothesisSource =
BaseLearner::RegisteredLearners().getLearner("ConstantLearner");
// get the training input data, and load it
InputData* pTrainingData = pWeakHypothesisSource->createInputData();
pTrainingData->initOptions(args);
pTrainingData->load(_trainFileName, IT_TRAIN, _verbose);
// get the testing input data, and load it
InputData* pTestData = NULL;
if ( !_testFileName.empty() )
{
pTestData = pWeakHypothesisSource->createInputData();
pTestData->initOptions(args);
pTestData->load(_testFileName, IT_TEST, _verbose);
}
// The output information object
OutputInfo* pOutInfo = NULL;
if ( !_outputInfoFile.empty() )
{
// Baseline: constant classifier - goes into 0th iteration
BaseLearner* pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal constantEnergy = pConstantWeakHypothesis->run();
//pOutInfo = new OutputInfo(_outputInfoFile);
pOutInfo = new OutputInfo(args);
pOutInfo->initialize(pTrainingData);
if (pTestData)
pOutInfo->initialize(pTestData);
pOutInfo->outputHeader(pTrainingData->getClassMap());
pOutInfo->outputIteration(-1);
pOutInfo->outputCustom(pTrainingData, pConstantWeakHypothesis);
if (pTestData != NULL)
{
pOutInfo->separator();
pOutInfo->outputCustom(pTestData, pConstantWeakHypothesis);
}
pOutInfo->outputCurrentTime();
pOutInfo->endLine();
pOutInfo->initialize(pTrainingData);
if (pTestData)
pOutInfo->initialize(pTestData);
}
//cout << "Before serialization" << endl;
// reload the previously found weak learners if -resume is set.
// otherwise just return 0
int startingIteration = resumeWeakLearners(pTrainingData);
Serialization ss(_shypFileName, _isShypCompressed );
ss.writeHeader(_baseLearnerName); // this must go after resumeProcess has been called
// perform the resuming if necessary. If not it will just return
resumeProcess(ss, pTrainingData, pTestData, pOutInfo);
if (_verbose == 1)
cout << "Learning in progress..." << endl;
//I put here the starting time, but it may take very long time to load the saved model
time_t startTime, currentTime;
time(&startTime);
///////////////////////////////////////////////////////////////////////
// Starting the AdaBoost main loop
///////////////////////////////////////////////////////////////////////
for (int t = startingIteration; t < _numIterations; ++t)
{
if (_verbose > 1)
cout << "------- WORKING ON ITERATION " << (t+1) << " -------" << endl;
BaseLearner* pWeakHypothesis = pWeakHypothesisSource->create();
pWeakHypothesis->initLearningOptions(args);
//pTrainingData->clearIndexSet();
pWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal energy = pWeakHypothesis->run();
//float gamma = pWeakHypothesis->getEdge();
//cout << gamma << endl;
if ( (_withConstantLearner) || ( energy != energy ) ) // check constant learner if user wants it (if energi is nan, then we chose constant learner
{
BaseLearner* pConstantWeakHypothesis = pConstantWeakHypothesisSource->create() ;
pConstantWeakHypothesis->initLearningOptions(args);
pConstantWeakHypothesis->setTrainingData(pTrainingData);
AlphaReal constantEnergy = pConstantWeakHypothesis->run();
if ( (constantEnergy <= energy) || ( energy != energy ) ) {
delete pWeakHypothesis;
pWeakHypothesis = pConstantWeakHypothesis;
}
}
if (_verbose > 1)
cout << "Weak learner: " << pWeakHypothesis->getName()<< endl;
// Output the step-by-step information
printOutputInfo(pOutInfo, t, pTrainingData, pTestData, pWeakHypothesis);
// Updates the weights and returns the edge
AlphaReal gamma = updateWeights(pTrainingData, pWeakHypothesis);
if (_verbose > 1)
{
cout << setprecision(5)
<< "--> Alpha = " << pWeakHypothesis->getAlpha() << endl
<< "--> Edge = " << gamma << endl
<< "--> Energy = " << energy << endl
// << "--> ConstantEnergy = " << constantEnergy << endl
// << "--> difference = " << (energy - constantEnergy) << endl
;
}
// If gamma <= theta the algorithm must stop.
// If theta == 0 and gamma is 0, it means that the weak learner is no better than chance
// and no further training is possible.
if (gamma <= _theta)
{
if (_verbose > 0)
{
cout << "Can't train any further: edge = " << gamma
<< " (with and edge offset (theta)=" << _theta << ")" << endl;
}
// delete pWeakHypothesis;
// break;
}
// append the current weak learner to strong hypothesis file,
// that is, serialize it.
ss.appendHypothesis(t, pWeakHypothesis);
// Add it to the internal list of weak hypotheses
_foundHypotheses.push_back(pWeakHypothesis);
// check if the time limit has been reached
if (_maxTime > 0)
{
time( ¤tTime );
float diff = difftime(currentTime, startTime); // difftime is in seconds
diff /= 60; // = minutes
if (diff > _maxTime)
{
if (_verbose > 0)
cout << "Time limit of " << _maxTime
<< " minutes has been reached!" << endl;
break;
}
} // check for maxtime
delete pWeakHypothesis;
} // loop on iterations
/////////////////////////////////////////////////////////
// write the footer of the strong hypothesis file
ss.writeFooter();
// write the weights of the instances if the name of weights file isn't empty
printOutWeights( pTrainingData );
// Free the two input data objects
if (pTrainingData)
delete pTrainingData;
if (pTestData)
delete pTestData;
if (pOutInfo)
delete pOutInfo;
if (_verbose > 0)
cout << "Learning completed." << endl;
}
// -----------------------------------------------------------------------
// -----------------------------------------------------------------------
double DataReader::getAccuracyOnCurrentDataSet()
{
double acc=0.0;
const int numClasses = _pCurrentData->getNumClasses();
const int numExamples = _pCurrentData->getNumExamples();
int correct=0;
int incorrect=0;
for( int i = 1; i < numExamples; i++ )
{
ExampleResults* tmpResult = new ExampleResults( i, numClasses );
vector<AlphaReal>& currVotesVector = tmpResult->getVotesVector();
for( int j=0; j<_weakHypotheses.size(); j++ )
{
BaseLearner* currWeakHyp = _weakHypotheses[j];
float alpha = currWeakHyp->getAlpha();
// for every class
for (int l = 0; l < numClasses; ++l)
currVotesVector[l] += alpha * currWeakHyp->classify(_pCurrentData, i, l);
}
vector<Label>::const_iterator lIt;
const vector<Label>& labels = _pCurrentData->getLabels(i);
// the vote of the winning negative class
float maxNegClass = -numeric_limits<float>::max();
// the vote of the winning positive class
float minPosClass = numeric_limits<float>::max();
for ( lIt = labels.begin(); lIt != labels.end(); ++lIt )
{
// get the negative winner class
if ( lIt->y < 0 && currVotesVector[lIt->idx] > maxNegClass )
maxNegClass = currVotesVector[lIt->idx];
// get the positive winner class
if ( lIt->y > 0 && currVotesVector[lIt->idx] < minPosClass )
minPosClass = currVotesVector[lIt->idx];
}
// if the vote for the worst positive label is lower than the
// vote for the highest negative label -> error
if (minPosClass <= maxNegClass)
incorrect++;
else {
correct++;
}
}
acc = ((double) correct / ((double) numExamples)) * 100.0;
return acc;
}