void output_histograms(sampler *samp, char matlab_hist, char gnuplot_hist){
/*
Compute histograms for the given sampler.
Number of bins in this script is always 100.
Bin locations are picked dynamically to not lose samples.
Input:
sampler *samp Sampler object with sampler already run.
char matlab_hist If true, will write matlab format data files.
char gnuplot_hist If true, will write gnuplot format data files.
Output:
Write data files for histograms.
*/
int n_bins = 100;
double tau;
float *centers = (float *) malloc(n_bins * sizeof(float));
if(!centers){ perror("Allocation failure Histogram"); abort(); }
float *f_hat = (float *) malloc(n_bins * sizeof(float));
if(!f_hat){ perror("Allocation failure Histogram"); abort(); }
float *X = (float *) malloc(samp->total_samples * sizeof(float));
if(!X){ perror("Allocation failure Histogram"); abort(); }
for(int i=0; i<samp->num_to_save; i++){
for(int j=0; j < samp->total_samples; j++)
X[j] = samp->samples_host[i + j*(samp->num_to_save)];
tau = samp->acor_times[i];
histogram_data(n_bins, X, samp->total_samples, tau, centers, f_hat);
if(matlab_hist)
histogram_to_matlab(n_bins, centers, f_hat, (samp->indices_to_save_host[i])+1); // add one for matlab index
if(gnuplot_hist)
histogram_to_gnuplot(n_bins, centers, f_hat, samp->indices_to_save_host[i]);
}
free(centers);
free(f_hat);
free(X);
}
void fit(TH1* histogram, double xMin, double xMax, double massValue, std::vector<fitParameterType>& fitParameter)
{
// create fit variable
TAxis* xAxis = histogram->GetXaxis();
if ( xMin < 0. ) xMin = xAxis->GetXmin();
if ( xMax < 0. ) xMax = xAxis->GetXmax();
std::string fitVariableName = Form("%s_fitVariable", histogram->GetName());
RooRealVar fitVariable(fitVariableName.data(), fitVariableName.data(), xMin, xMax);
// convert histogram to RooFit format
std::string histogramName_data = Form("%s_data", histogram->GetName());
RooDataHist histogram_data(histogramName_data.data(), histogramName_data.data(), fitVariable, histogram, 1.0);
// create product of linear * Heaviside * Gaussian Error function
std::string svFitLineShapeName_slope = Form("%s_svFitLineShape_slope", histogram->GetName());
RooRealVar svFitLineShape_slope(svFitLineShapeName_slope.data(), svFitLineShapeName_slope.data(), 2./(massValue*massValue), 0., 1.);
std::string svFitLineShapeName_offset = Form("%s_svFitLineShape_offset", histogram->GetName());
RooRealVar svFitLineShape_offset(svFitLineShapeName_offset.data(), svFitLineShapeName_offset.data(), 0., -1., +1.);
std::string heavisideName_threshold = Form("%s_heaviside_threshold", histogram->GetName());
RooRealVar heaviside_threshold(heavisideName_threshold.data(), heavisideName_threshold.data(), 0.2*massValue, 1.e-2*massValue, 0.5*massValue);
std::string sculptingName_bias = Form("%s_sculpting_bias", histogram->GetName());
RooRealVar sculpting_bias(sculptingName_bias.data(), sculptingName_bias.data(), 0.25*massValue, 0.*massValue, 0.9*massValue);
std::string sculptingName_width = Form("%s_sculpting_width", histogram->GetName());
RooRealVar sculpting_width(sculptingName_width.data(), sculptingName_width.data(), 0.1*massValue, 1.e-2*massValue, 1.0*massValue);
std::string svFitLineShapeName = Form("%s_svFitLineShape", histogram->GetName());
std::string svFitLineShapeFormula = "(@0*@1 + @2)*0.5*(1.0 + TMath::Sign(+1, @0 - @3))*0.5*(1.0 + TMath::Erf((@0 - @4)/@5))";
RooArgList svFitLineShapeArgs(fitVariable,
svFitLineShape_slope,
svFitLineShape_offset,
heaviside_threshold,
sculpting_bias,
sculpting_width);
RooGenericPdf svFitLineShape(svFitLineShapeName.data(), svFitLineShapeName.data(), svFitLineShapeFormula.data(), svFitLineShapeArgs);
// create Gaussian
std::string gaussianName_mean = Form("%s_gaussian_mean", histogram->GetName());
RooConstVar gaussian_mean(gaussianName_mean.data(), gaussianName_mean.data(), 0.);
std::string gaussianName_sigma = Form("%s_gaussian_sigma", histogram->GetName());
RooRealVar gaussian_sigma(gaussianName_sigma.data(), gaussianName_sigma.data(), 0.2*massValue, 1.e-2*massValue, 0.5*massValue);
std::string gaussianName = Form("%s_gaussian", histogram->GetName());
RooGaussian gaussian(gaussianName.data(), gaussianName.data(), fitVariable, gaussian_mean, gaussian_sigma);
// numerically convolute both PDFs using fast Fourier transform
std::string fitFunctionName = Form("%s_fitFunction", histogram->GetName());
RooFFTConvPdf fitFunction(fitFunctionName.data(), fitFunctionName.data(), fitVariable, svFitLineShape, gaussian);
// fit histogram
fitFunction.fitTo(histogram_data);
// save fit parameter
fitParameter.resize(6);
fitParameter[0] = fitParameterType(&svFitLineShape_slope);
fitParameter[1] = fitParameterType(&svFitLineShape_offset);
fitParameter[2] = fitParameterType(&gaussian_sigma);
fitParameter[3] = fitParameterType(&heaviside_threshold);
fitParameter[4] = fitParameterType(&sculpting_bias);
fitParameter[5] = fitParameterType(&sculpting_width);
// create control plot
std::string frameTitle = Form("%s_frame", histogram->GetName());
RooPlot* frame = fitVariable.frame(RooFit::Title(frameTitle.data()));
histogram_data.plotOn(frame);
fitFunction.plotOn(frame, RooFit::LineColor(kRed));
std::string canvasName = Form("%s_canvas", histogram->GetName());
TCanvas* canvas = new TCanvas(canvasName.data(), canvasName.data(), 800, 600);
gPad->SetLeftMargin(0.15);
frame->GetYaxis()->SetTitleOffset(1.4);
frame->Draw();
std::string outputFileName_plot = Form("svFitPerformance_%s_fit", histogram->GetName());
canvas->Print(std::string(outputFileName_plot).append(".eps").data());
canvas->Print(std::string(outputFileName_plot).append(".png").data());
canvas->Print(std::string(outputFileName_plot).append(".pdf").data());
delete canvas;
}
virtual bool run()
{
const std::vector<sd::core::ViewNode *> &parents = this->getPreviousViewNodes();
sd::core::ImageView_<T> *image = static_cast<sd::core::ImageView_<T>*>(parents[0]);
if (parents.size() != 1)
{
std::cout << "\tExpecting 1 parent exactly, found " << parents.size() << ". Abort!\n";
return false;
}
sd::frontend::Parameter pchannel;
this->getParams("channel", pchannel);
std::string wanted_channel = pchannel.getString();
std::cout << "SELECTED channel: " << wanted_channel << std::endl;
sd::frontend::Parameter pNbTiles;
this->getParams("nbTiles", pNbTiles);
int nbTiles = pNbTiles.getInteger();
/* RGBA: R is 0, G is 1...*/
int channel_id = 0;
if(wanted_channel == "Red")
channel_id = 0;
else if(wanted_channel == "Green")
channel_id = 1;
else if(wanted_channel == "Blue")
channel_id = 2;
else if(wanted_channel == "Alpha")
channel_id = 3;
else
std::cout << "Error on channel, default: red" << std::endl;
std::vector<float> histogram_data_1(256, 0);
std::vector<float> histogram_data(nbTiles, 0);
int loop_count = 0;
int value = 0;
for(auto source_it = image->begin(); source_it != image->end(); ++source_it){
value = (float)source_it(channel_id);
histogram_data_1[value]++;
loop_count++;
}
int coef = 256/nbTiles;
for(int i=0; i<nbTiles; i++){
for(int j=0; j<coef; j++){
histogram_data[i] += histogram_data_1[(i+1)*coef+j];
}
}
float maxHeight = 0;
for(int i=0; i<=nbTiles; i++){
if(maxHeight < histogram_data[i]){
maxHeight = histogram_data[i];
}
}
if(maxHeight != 0){
for(int i=0; i<=nbTiles; i++){
histogram_data[i] = (histogram_data[i]*256)/maxHeight;
}
}
int sizeX = nbTiles;
int sizeY = nbTiles;
sd::Size result_size(sizeX, sizeY);
sd::core::ImageViewInfo info(sd::GRAYLEVEL, sd::Z_AXIS, result_size);
this->init(info);
T* resultData = this->getData();
std::fill(resultData, resultData+result_size.dataSize(), 255);
this->setMinMax(0, 255);
auto itR = this->begin();
for(int i = 0; i < nbTiles; i++){
for(int j = 0; j < nbTiles; j++, itR++){
for(int k = 0; k < 4; k++){
if((i) < (int)(nbTiles-histogram_data[j]))
itR(k) = 255;
else
itR(k) = 0;
}
}
}
return true;
}
