void FidoInterface::gridSearchOptimize() {
if (VERB > 1) {
std::cerr << "Running super grid search..." << std::endl;
}
double alpha_step = 0.05;
double beta_step = 0.05;
double gamma_step = 0.05;
double beta_init = 0.00001;
double alpha_init = 0.001;
double gamma_init = 0.1;
double gamma_limit = 0.5;
double beta_limit = 0.05;
double alpha_limit = 0.5;
if (alpha_ != -1) alpha_init = alpha_limit = alpha_;
if ( beta_ != -1) beta_init = beta_limit = beta_;
if (gamma_ != -1) gamma_init = gamma_limit = gamma_;
std::vector<double> gamma_search, beta_search, alpha_search;
//NOTE very annoying the residue error of the floats that get acummulated in every iteration
for (int i = 0; gamma_init + gamma_step*i <= gamma_limit; ++i) {
gamma_search.push_back(gamma_init + gamma_step*i);
}
double lbi = log10(beta_init);
for (int i = 0; lbi + i*beta_step <= Round(log10(beta_limit),2); ++i) {
double j = lbi + i*beta_step;
double original = pow(10,j);
double beta_local = original - beta_init;
if (beta_local > 0.0) beta_local = original;
beta_search.push_back(beta_local);
}
double lai = log10(alpha_init);
for (int i = 0; lai + i*alpha_step <= Round(log10(alpha_limit),2); ++i) {
double k = lai + i*alpha_step;
alpha_search.push_back(pow(10,k));
}
gridSearch(alpha_search, beta_search, gamma_search);
}
int main(){
dic = inputDictionary();
int testCase, i, j, k, l, m, n, o, p;
char key;
scanf("%d\n", &testCase);
for(i=0; i<testCase; i++){
scanf("%d %d\n", &row, &column);
char line[column];
grid = malloc(sizeof(char*)*row);
for(j=0; j<row; j++){
grid[j] = malloc(sizeof(char)*column);
}
for(k=0; k<row; k++){
scanf("%s", line);
for(l=0; l<column; l++){
grid[k][l]= line[l];
}
}
printf("Words Found Gird #%d:\n", i+1);
gridSearch();
}
//After everything is done free the memory
for(i=0; i<row; i++){ //free dic
free(grid[i]);
}
free(grid);
for(i=0; i<num; i++){ //free dic
free(dic[i]);
}
free(dic);
return 0;
}
void FidoInterface::gridSearch() {
std::vector<double> gamma_search, beta_search, alpha_search;
switch(gridSearchDepth_) {
case 0:
gamma_search.push_back(0.5);
beta_search.push_back(0.001);
alpha_search.push_back(0.008);
alpha_search.push_back(0.032);
alpha_search.push_back(0.128);
break;
case 1:
gamma_search.push_back(0.1);
gamma_search.push_back(0.5);
gamma_search.push_back(0.9);
beta_search.push_back(0.001);
for (double k = 0.002; k <= 0.4; k*=4) {
alpha_search.push_back(k);
}
break;
case 2:
gamma_search.push_back(0.1);
gamma_search.push_back(0.3);
gamma_search.push_back(0.5);
gamma_search.push_back(0.7);
gamma_search.push_back(0.9);
beta_search.push_back(0.001);
for (double k = 0.001; k <= 0.4; k*=2) {
alpha_search.push_back(k);
}
break;
case 3:
for (double k = 0.01; k <= 0.76; k+=0.05) {
alpha_search.push_back(k);
}
beta_search.push_back(0.001);
for (double k = 0.05; k <= 0.95; k+=0.05) {
gamma_search.push_back(k);
}
break;
case 4: // grid search from FIDO paper
for (double k = 0.01; k <= 0.76; k+=0.05) {
alpha_search.push_back(k);
}
for (double k = 0.0; k <= 0.80; k+=0.05) {
beta_search.push_back(k);
}
for (double k = 0.1; k <= 0.9; k+=0.1) {
gamma_search.push_back(k);
}
break;
default:
gamma_search.push_back(0.5);
beta_search.push_back(0.001);
alpha_search.push_back(0.008);
alpha_search.push_back(0.032);
alpha_search.push_back(0.128);
break;
}
if (alpha_ != -1) alpha_search.push_back(alpha_);
if ( beta_ != -1) beta_search.push_back(beta_);
if (gamma_ != -1) gamma_search.push_back(gamma_);
gridSearch(alpha_search, beta_search, gamma_search);
}
void FidoInterface::computeProbabilities(const std::string& fname) {
ifstream fin;
if (fname.size() > 0) {
fin.open(fname.c_str());
proteinGraph_->read(fin);
} else {
proteinGraph_->read(peptideScores_);
}
if (trivialGrouping_) updateTargetDecoySizes();
time_t startTime;
clock_t startClock;
time(&startTime);
startClock = clock();
if (mayufdr) {
computeFDR();
}
if (doGridSearch_) {
if (VERB > 1) {
std::cerr << "The parameters for the model will be estimated by grid search.\n" << std::endl;
}
if (kOptimizeParams)
gridSearchOptimize();
else
gridSearch();
time_t procStart;
clock_t procStartClock = clock();
time(&procStart);
double diff = difftime(procStart, startTime);
if (VERB > 1) cerr << "Estimating the parameters took : "
<< ((double)(procStartClock - startClock)) / (double)CLOCKS_PER_SEC
<< " cpu seconds or " << diff << " seconds wall time" << endl;
}
if (VERB > 1) {
cerr << "The following parameters have been chosen:\n";
std::cerr.precision(10);
cerr << "alpha = " << alpha_ << endl;
cerr << "beta = " << beta_ << endl;
cerr << "gamma = " << gamma_ << endl;
std::cerr.unsetf(std::ios::floatfield);
cerr << "\nProtein level probabilities will now be estimated\n";
}
if (gridSearchThreshold_ > 0.0 && doGridSearch_) {
//NOTE reset the tree after grid searching
proteinGraph_->setProteinThreshold(proteinThreshold_);
proteinGraph_->setPsmThreshold(kPsmThreshold);
proteinGraph_->setPeptideThreshold(kPeptideThreshold);
proteinGraph_->setNoClustering(noClustering_);
proteinGraph_->setNoPartitioning(noPartitioning_);
proteinGraph_->setNoPruning(noPruning_);
proteinGraph_->setTrivialGrouping(trivialGrouping_);
proteinGraph_->setMultipleLabeledPeptides(kAddPeptideDecoyLabel);
if (fname.size() > 0) {
proteinGraph_->read(fin);
} else {
proteinGraph_->read(peptideScores_);
}
if (trivialGrouping_) updateTargetDecoySizes();
}
proteinGraph_->setAlphaBetaGamma(alpha_, beta_, gamma_);
proteinGraph_->getProteinProbs();
pepProteinMap_.clear();
proteinGraph_->getProteinProbsPercolator(pepProteinMap_);
}
//NOTE almost entirely duplicated of computeProbabilities, it could be refactored
void FidoInterface::computeProbabilitiesFromFile(ifstream &fin)
{
proteinGraph->read(fin);
time_t startTime;
clock_t startClock;
time(&startTime);
startClock = clock();
if(mayufdr)
{
computeFDR();
}
if(dogridSearch)
{
if(VERB > 1)
{
std::cerr << "The parameters for the model will be estimated by grid search.\n" << std::endl;
}
if(optimize)
gridSearchOptimize();
else
gridSearch();
time_t procStart;
clock_t procStartClock = clock();
time(&procStart);
double diff = difftime(procStart, startTime);
if (VERB > 1) cerr << "Estimating the parameters took : "
<< ((double)(procStartClock - startClock)) / (double)CLOCKS_PER_SEC
<< " cpu seconds or " << diff << " seconds wall time" << endl;
}
if(VERB > 1)
{
cerr << "The following parameters have been chosen:\n";
std::cerr.precision(10);
cerr << "gamma = " << gamma << endl;
cerr << "alpha = " << alpha << endl;
cerr << "beta = " << beta << endl;
std::cerr.unsetf(std::ios::floatfield);
cerr << "\nProtein level probabilities will now be estimated";
}
if(dogridSearch && reduceTree)
{
//NOTE lets create the tree again with all the members
double local_protein_threshold = proteinThreshold;
if(truncate) local_protein_threshold = 0.0;
proteinGraph->setProteinThreshold(local_protein_threshold);
proteinGraph->setPsmThreshold(psmThreshold);
proteinGraph->setPeptideThreshold(peptideThreshold);
proteinGraph->setGroupProteins(nogroupProteins);
proteinGraph->setSeparateProteins(noseparate);
proteinGraph->setPruneProteins(noprune);
proteinGraph->setMultipleLabeledPeptides(allow_multiple_labeled_peptides);
proteinGraph->read(fin);
}
proteinGraph->setAlphaBetaGamma(alpha,beta,gamma);
proteinGraph->getProteinProbs();
pepProteins.clear();
proteinGraph->getProteinProbsPercolator(pepProteins);
}
