Filter::Filter(InputSet& inputSet) : inputSet_(inputSet) {
// Get the input set's number of samples, features and labels
const unsigned int nbSamples = inputSet.nbSamples();
const unsigned int nbFeatures = inputSet.nbFeatures();
const unsigned int nbLabels = inputSet.nbLabels();
// Resize the initial stack of samples
sampleStack_.resize(1);
sampleStack_[0].resize(nbSamples);
// Make the indices range from 0 to nbSamples - 1
for(unsigned int s = 0; s < nbSamples; ++s) {
sampleStack_[0][s] = s;
}
// Resize the initial stack of features
featureStack_.resize(1);
featureStack_[0].resize(nbFeatures);
// Make the indices range from 0 to nbFeatures - 1
for(unsigned int f = 0; f < nbFeatures; ++f) {
featureStack_[0][f] = f;
}
// Set the number of labels
nbLabelStack_.push_back(nbLabels); // Push it twice so as to be sure to
nbLabelStack_.push_back(nbLabels); // never overwrite it
// Set the number of images and heuristics
nbImages_ = inputSet.nbImages();
nbHeuristics_ = inputSet.nbHeuristics();
}
Statistical::Statistical(InputSet& inputSet) : Filter(inputSet) {
// Get the number of samples and features
const unsigned int nbSamples = inputSet.nbSamples();
const unsigned int nbFeatures = inputSet.nbFeatures();
// Used to store the current sum and sum of squares of every feature
means_.resize(nbFeatures);
stds_.resize(nbFeatures);
for(unsigned int s = 0; s < nbSamples; ++s) {
std::vector<unsigned int> sample(1, s);
inputSet.pushSamples(sample);
const scalar_t* features = inputSet.features(0);
for(unsigned int f = 0; f < nbFeatures; ++f) {
means_[f] += features[f];
stds_[f] += features[f] * features[f];
}
inputSet.popSamples();
}
std::vector<unsigned int> featureStack;
for(unsigned int f = 0; f < nbFeatures; ++f) {
scalar_t mean = means_[f] / nbSamples;
scalar_t variance = (stds_[f] - means_[f] * mean) / (nbSamples - 1);
if(variance > 0) {
means_[f] = mean;
stds_[f] = std::sqrt(variance);
featureStack.push_back(f);
}
}
// There must be at least one feature
assert(!featureStack.empty());
// Push the selected features if needed
if(featureStack.size() < nbFeatures) {
featureStack_.push_back(featureStack);
}
}
void C45Tree::train(InputSet& inputSet) {
// In case the tree was already trained
delete children_[0];
delete children_[1];
children_[0] = 0;
children_[1] = 0;
// Get the number of samples and labels
const unsigned int nbSamples = inputSet.nbSamples();
const unsigned int nbLabels = inputSet.nbLabels();
// Set the maximal depth if needed
bool maxDepth = false;
if(!maxDepth_) {
maxDepth_ = std::ceil(std::log(double(nbLabels)) / std::log(2.0));
maxDepth = true;
}
// Get the labels and the weights of the samples
const unsigned int* labels = inputSet.labels();
const scalar_t* weights = inputSet.weights();
// Make the weights sum to nbSamples as in Dr. Quilans implementation
std::vector<scalar_t> oldWeights(nbSamples);
scalar_t norm = std::accumulate(weights, weights + nbSamples, scalar_t());
std::transform(weights, weights + nbSamples, oldWeights.begin(),
std::bind2nd(std::multiplies<scalar_t>(), nbSamples / norm));
inputSet.swapWeights(oldWeights);
weights = inputSet.weights();
// Compute the frequency of appearance of each label
distr_.clear();
distr_.resize(nbLabels, 0);
// Determine the label with the largest frequency
for(unsigned int s = 0; s < nbSamples; ++s) {
distr_[labels[s]] += weights[s];
}
label_ = std::max_element(distr_.begin(), distr_.end()) - distr_.begin();
sumWeights_ = nbSamples;
// Vector of indices over the samples
std::vector<unsigned int> indices(nbSamples);
// Make the indices range from o to nbSamples - 1
for(unsigned int s = 0; s < nbSamples; ++s) {
indices[s] = s;
}
// Create the root (the children will follow recursively)
make(inputSet, &indices[0], nbSamples);
if(confidence_ < 1) {
// Make the indices range from o to nbSamples - 1
for(unsigned int s = 0; s < nbSamples; ++s) {
indices[s] = s;
}
// Prune the tree previously created
prune(inputSet, &indices[0], nbSamples, true);
}
// Restore the original weights
inputSet.swapWeights(oldWeights);
// Restore maxDepth
if(maxDepth) {
maxDepth_ = 0;
}
}
void LinearSVM::train(InputSet& inputSet) {
// Sample features from every heuristic in order for the matrix of features
// to fit in memory
if(inputSet.nbFeatures() > NB_FEATURES_MAX / inputSet.nbSamples()) {
inputSet.sampleFeatures(NB_FEATURES_MAX / inputSet.nbSamples(), indices_);
inputSet.pushFeatures(indices_);
}
else {
indices_.clear();
}
// Get the number of features, samples and labels
const unsigned int nbFeatures = inputSet.nbFeatures();
const unsigned int nbSamples = inputSet.nbSamples();
const unsigned int nbLabels = inputSet.nbLabels();
// Delete the previous model
if(model_) {
free_and_destroy_model(&model_);
}
// Create a new problem
problem prob;
// Recopy the number of samples
prob.l = nbSamples;
prob.n = nbFeatures;
// Recopy the labels
std::vector<int> labels(inputSet.labels(), inputSet.labels() + nbSamples);
prob.y = &labels[0];
// Recopy the features (as we want to normalize them)
std::vector<scalar_t> matrix;
inputSet.swapFeatures(matrix);
// Create samples as expected by liblinear
std::vector<feature_node> samples(nbSamples);
prob.x = &samples[0];
// Compute the mean norm
scalar_t meanNorm = 0;
for(unsigned int s = 0; s < nbSamples; ++s) {
samples[s].dim = nbFeatures;
samples[s].values = &matrix[s * nbFeatures];
// Add the mean norm of that sample
meanNorm += nrm2(samples[s].dim, samples[s].values, 1);
}
// Divide the sum of the norms by the number of samples
meanNorm /= nbSamples;
std::cout << "[LinearSVM::train] mean(norm): " << meanNorm << '.'
<< std::endl;
// Rescale the features so that their mean norm is 1
std::transform(matrix.begin(), matrix.end(), matrix.begin(),
std::bind2nd(std::divides<scalar_t>(), meanNorm));
// Sets the bias to the default value (liblinear doesn't seem to handle the
// bias parameter value correctly)
prob.bias = -1;
// Make sure that the parameters are correct
bool crossValidate = parameters_.C < 0;
// Sets C to a default value in order to pass the parameter check
if(crossValidate) {
parameters_.C = 1;
}
// There is a problem with the parameters
assert(!check_parameter(&prob, ¶meters_));
// If C is below zero, use 5-folds cross-validation to determine it
if(crossValidate) {
std::vector<int> target(nbSamples); // The predicted labels
unsigned int nbErrorsMin = nbSamples + 1; // Initialize past the maximum
for(parameters_.C = 1000; parameters_.C >= 0.01; parameters_.C /= 10) {
cross_validation(&prob, ¶meters_, 5, &target[0]);
// Count the number of errors
unsigned int nbErrors = 0;
for(unsigned int s = 0; s < nbSamples; ++s) {
if(target[s] != labels[s]) {
++nbErrors;
}
}
std::cout << "[LinearSVM::train] 5 folds cross-validation error "
"for C = " << parameters_.C << ": "
<< nbErrors * 100.0f / nbSamples << "%." << std::endl;
// The new C is better than the previous one
if(nbErrors < nbErrorsMin) {
nbErrorsMin = nbErrors;
}
// The optimal C was found
else {
break;
}
}
// C got divided one time too much
parameters_.C *= 10;
// Print C to the log
std::cout << "[LinearSVM::train] optimal C as determined by 5 folds "
"cross-validation: " << parameters_.C << '.' << std::endl;
}
// Train the svm
model_ = ::train(&prob, ¶meters_);
assert(model_);
// Reset C so that it will be cross-validated again
if(crossValidate) {
parameters_.C = -1;
}
// Save libsvm labels
map_.clear();
map_.resize(nbLabels);
get_labels(model_, &map_[0]);
// Pop the selected features if required
if(!indices_.empty()) {
inputSet.popFeatures();
}
}
