// Alpha-nDCG Computation
static arma::rowvec compute_andcg_discount(arma::rowvec subtopics, double alpha){
// Iterate through the elements in the vector and compute (1- alpha)^x
arma::rowvec discount = arma::zeros<arma::rowvec>(subtopics.n_cols);
int i =0;
for(arma::rowvec::iterator it=subtopics.begin(); it!=subtopics.end(); ++it){
discount(i++) = pow((1 - alpha),(*it));
}
return discount;
}
/** For debugging reasons*/
void checkAllComponents() {
arma::vec rowSumA = arma::sum(A_, 1);
rowSumA.print("rowSumA");
double sumPi = arma::sum(pi_);
std::cout << "sumPi: " << sumPi << std::endl;
arma::vec weights = arma::zeros((unsigned int)BModels_.size());
for (unsigned int i = 0; i < (unsigned int) BModels_.size(); ++i) {
weights(i) = arma::accu(BModels_[i].getWeights());
}
weights.print("bCumWeights");
arma::rowvec checksum = arma::sum(gamma_);
checksum.print("checksum");
arma::uvec checksumIndices = arma::find(checksum < 1.0 - 1E-2);
if (checksumIndices.n_elem >= 1) {
arma::rowvec checksumAlpha = arma::sum(alpha_);
checksumAlpha.print("checkAlpha");
//alpha_.print("alpha");
arma::rowvec checksumBeta = arma::sum(beta_);
checksumBeta.print("checkBeta");
//beta_.print("beta");
c_.print("c");
throw std::runtime_error("data going wonky");
}
if (!arma::is_finite(A_)) {
A_.print("A Fail");
throw std::runtime_error("A has invalid entries");
}
if (!arma::is_finite(pi_)) {
pi_.print("pi Fail");
throw std::runtime_error("pi has invalid entries");
}
if (!arma::is_finite(alpha_)) {
alpha_.print("alpha Fail");
throw std::runtime_error("alpha has invalid entries");
}
if (!arma::is_finite(beta_)) {
beta_.print("beta Fail");
throw std::runtime_error("beta has invalid entries");
}
if (!arma::is_finite(gamma_)) {
gamma_.print("gamma Fail");
throw std::runtime_error("gamma has invalid entries");
}
if (!arma::is_finite(xi_)) {
xi_.print("xi Fail");
throw std::runtime_error("xi has invalid entries");
}
}
void
print(std::string header = "") {
A_.print("A");
for (size_t i = 0; i < BModels_.size(); ++i) {
BModels_[i].print("B");
}
pi_.print("pi");
}
void Adaboost<MatType, WeakLearner>::buildWeightMatrix(
const arma::mat& D, arma::rowvec& weights)
{
int i, j;
weights.fill(0.0);
for (i = 0;i < D.n_rows; i++)
{
for (j = 0;j < D.n_cols; j++)
weights(i) += D(i,j);
}
}
void pca::read_armvct (const char * fname, arma::rowvec & q)
{
int n;
std::ifstream myfileq(fname, std::ios::binary);
myfileq.read((char*)&(n), sizeof(n));
q.set_size(n);
for (int i=0; i<(int)q.n_cols; i++)
{
double v;
myfileq.read((char*)&v, sizeof(v));
q(i) = v;
}
myfileq.close();
}
double DecisionStump<MatType>::SetupSplitDimension(
const arma::rowvec& dimension,
const arma::Row<size_t>& labels,
const arma::rowvec& weights)
{
size_t i, count, begin, end;
double entropy = 0.0;
// Sort the dimension in order to calculate splitting ranges.
arma::rowvec sortedDim = arma::sort(dimension);
// Store the indices of the sorted dimension to build a vector of sorted
// labels. This sort is stable.
arma::uvec sortedIndexDim = arma::stable_sort_index(dimension.t());
arma::Row<size_t> sortedLabels(dimension.n_elem);
arma::rowvec sortedWeights(dimension.n_elem);
for (i = 0; i < dimension.n_elem; i++)
{
sortedLabels(i) = labels(sortedIndexDim(i));
// Apply weights if necessary.
if (UseWeights)
sortedWeights(i) = weights(sortedIndexDim(i));
}
i = 0;
count = 0;
// This splits the sorted data into buckets of size greater than or equal to
// bucketSize.
while (i < sortedLabels.n_elem)
{
count++;
if (i == sortedLabels.n_elem - 1)
{
// If we're at the end, then don't worry about the bucket size; just take
// this as the last bin.
begin = i - count + 1;
end = i;
// Use ratioEl to calculate the ratio of elements in this split.
const double ratioEl = ((double) (end - begin + 1) / sortedLabels.n_elem);
entropy += ratioEl * CalculateEntropy<UseWeights>(
sortedLabels.subvec(begin, end), sortedWeights.subvec(begin, end));
i++;
}
else if (sortedLabels(i) != sortedLabels(i + 1))
{
// If we're not at the last element of sortedLabels, then check whether
// count is less than the current bucket size.
if (count < bucketSize)
{
// If it is, then take the minimum bucket size anyways.
// This is where the inpBucketSize comes into use.
// This makes sure there isn't a bucket for every change in labels.
begin = i - count + 1;
end = begin + bucketSize - 1;
if (end > sortedLabels.n_elem - 1)
end = sortedLabels.n_elem - 1;
}
else
{
// If it is not, then take the bucket size as the value of count.
begin = i - count + 1;
end = i;
}
const double ratioEl = ((double) (end - begin + 1) / sortedLabels.n_elem);
entropy += ratioEl * CalculateEntropy<UseWeights>(
sortedLabels.subvec(begin, end), sortedWeights.subvec(begin, end));
i = end + 1;
count = 0;
}
else
i++;
}
return entropy;
}
double mahalanobis(const arma::rowvec& x, const arma::rowvec& mu, const arma::mat& sigma) {
const arma::rowvec err = x - mu;
return arma::as_scalar(err * sigma.i() * err.t());
}
double mahalanobis_chol(const arma::rowvec& x, const arma::rowvec& mu, const arma::mat& R) {
const arma::rowvec err = x - mu;
const arma::mat Rinv(inv(trimatl(R)));
return arma::as_scalar(err * Rinv * Rinv.t() * err.t());
}
void DecisionStump<MatType>::TrainOnAtt(const arma::rowvec& attribute,
const arma::Row<size_t>& labels)
{
size_t i, count, begin, end;
arma::rowvec sortedSplitAtt = arma::sort(attribute);
arma::uvec sortedSplitIndexAtt = arma::stable_sort_index(attribute.t());
arma::Row<size_t> sortedLabels(attribute.n_elem);
sortedLabels.fill(0);
arma::vec tempSplit;
arma::Row<size_t> tempLabel;
for (i = 0; i < attribute.n_elem; i++)
sortedLabels(i) = labels(sortedSplitIndexAtt(i));
arma::rowvec subCols;
rType mostFreq;
i = 0;
count = 0;
while (i < sortedLabels.n_elem)
{
count++;
if (i == sortedLabels.n_elem - 1)
{
begin = i - count + 1;
end = i;
arma::rowvec zSubCols((sortedLabels.cols(begin, end)).n_elem);
zSubCols.fill(0.0);
subCols = sortedLabels.cols(begin, end) + zSubCols;
mostFreq = CountMostFreq<double>(subCols);
split.resize(split.n_elem + 1);
split(split.n_elem - 1) = sortedSplitAtt(begin);
binLabels.resize(binLabels.n_elem + 1);
binLabels(binLabels.n_elem - 1) = mostFreq;
i++;
}
else if (sortedLabels(i) != sortedLabels(i + 1))
{
if (count < bucketSize)
{
// Test for different values of bucketSize, especially extreme cases.
begin = i - count + 1;
end = begin + bucketSize - 1;
if (end > sortedLabels.n_elem - 1)
end = sortedLabels.n_elem - 1;
}
else
{
begin = i - count + 1;
end = i;
}
arma::rowvec zSubCols((sortedLabels.cols(begin, end)).n_elem);
zSubCols.fill(0.0);
subCols = sortedLabels.cols(begin, end) + zSubCols;
// Find the most frequent element in subCols so as to assign a label to
// the bucket of subCols.
mostFreq = CountMostFreq<double>(subCols);
split.resize(split.n_elem + 1);
split(split.n_elem - 1) = sortedSplitAtt(begin);
binLabels.resize(binLabels.n_elem + 1);
binLabels(binLabels.n_elem - 1) = mostFreq;
i = end + 1;
count = 0;
}
else
i++;
}
// Now trim the split matrix so that buckets one after the after which point
// to the same classLabel are merged as one big bucket.
MergeRanges();
}