Likelihood.cpp

Go to the documentation of this file.

/********************************************************************
 *  Copyright (C) 2014 by Federico Marulli and Alfonso Veropalumbo  *
 *  federico.marulli3@unibo.it                                      *
 *                                                                  *
 *  This program is free software; you can redistribute it and/or   *
 *  modify it under the terms of the GNU General Public License as  *
 *  published by the Free Software Foundation; either version 2 of  *
 *  the License, or (at your option) any later version.             *
 *                                                                  *
 *  This program is distributed in the hope that it will be useful, *
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of  *
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the   *
 *  GNU General Public License for more details.                    *
 *                                                                  *
 *  You should have received a copy of the GNU General Public       *
 *  License along with this program; if not, write to the Free      *
 *  Software Foundation, Inc.,                                      *
 *  59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.       *
 ********************************************************************/
 
#include "LikelihoodParameters.h"
#include "Likelihood.h"
 
using namespace std;
 
using namespace cbl;
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::m_set_grid_likelihood_1D (const int npoints, const std::vector<std::vector<double>> parameter_limits, const std::string output_file)
{
  vector<double> pp(npoints), log_ll(npoints);
  double binSize = (parameter_limits[0][1]-parameter_limits[0][0])/(npoints-1);
  for (int i=0; i<npoints; i++) {
    pp[i] = parameter_limits[0][0]+binSize*i;
    vector<double> par = {pp[i]};
    log_ll[i] = this->log(par);
  }
 
  shared_ptr<statistics::STR_likelihood_inputs> likelihood_inputs = static_pointer_cast<statistics::STR_likelihood_inputs>(m_likelihood_inputs);
  likelihood_inputs->interp_function1D = make_shared<glob::FuncGrid>(glob::FuncGrid(pp, log_ll, "Spline"));
  
  m_log_likelihood_function_grid = LogLikelihood_1D_interpolated;
  m_likelihood_function_grid = [&] (vector<double> &par, const shared_ptr<void> input) { return exp(m_log_likelihood_function_grid(par, input));};
 
  if (output_file!= par::defaultString) {
    ofstream fout(output_file); checkIO(fout, output_file);
    for (int i=0; i<npoints; i++)
      fout << pp[i] << "  " << log_ll[i] << endl;
    
    fout.close();
  }
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::m_set_grid_likelihood_2D (const int npoints, const std::vector<std::vector<double>> parameter_limits, const std::string output_file)
{
  vector<double> pp1(npoints), pp2(npoints);
  vector<vector<double>> log_ll(npoints, vector<double>(npoints, 0));
  double binSize1 = (parameter_limits[0][1]-parameter_limits[0][0])/(npoints-1);
  double binSize2 = (parameter_limits[1][1]-parameter_limits[1][0])/(npoints-1);
  for (int i=0; i<npoints; i++) {
    pp1[i] = parameter_limits[0][0]+binSize1*i;
    for (int j=0; j<npoints; j++) {
      pp2[j] = parameter_limits[1][0]+binSize2*j;
      vector<double> par = {pp1[i], pp2[j]};
      log_ll[i][j] = this->log(par);
    }
  }
 
  shared_ptr<statistics::STR_likelihood_inputs> likelihood_inputs = static_pointer_cast<statistics::STR_likelihood_inputs>(m_likelihood_inputs);
  likelihood_inputs->interp_function2D = make_shared<glob::FuncGrid2D>(glob::FuncGrid2D(pp1, pp2, log_ll, "Spline"));
 
  m_log_likelihood_function_grid = LogLikelihood_2D_interpolated;
  m_likelihood_function_grid = [&] (vector<double> &par, const shared_ptr<void> input) { return exp(m_log_likelihood_function_grid(par, input));};
 
  if (output_file!= par::defaultString) {
    ofstream fout(output_file); checkIO(fout, output_file);
    for (int i=0; i<npoints; i++) {
      for (int j=0; j<npoints; j++)
    fout << pp1[i] << "  " << pp2[j] << " " << log_ll[i][j] << endl;
      fout << endl;
    }
    
    fout.close();
  }
 
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::m_set_grid_likelihood_1D (const std::string input_file)
{
  vector<double> pp, log_ll;
 
  read_vector(input_file, pp, log_ll);
 
  shared_ptr<statistics::STR_likelihood_inputs> likelihood_inputs = static_pointer_cast<statistics::STR_likelihood_inputs>(m_likelihood_inputs);
  likelihood_inputs->interp_function1D = make_shared<glob::FuncGrid>(glob::FuncGrid(pp, log_ll, "Spline"));
 
  m_log_likelihood_function_grid = LogLikelihood_1D_interpolated;
  m_likelihood_function_grid = [&] (vector<double> &par, const shared_ptr<void> input) { return exp(m_log_likelihood_function_grid(par, input));};
 
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::m_set_grid_likelihood_2D (const std::string input_file)
{
  vector<double> pp1, pp2;
  vector<vector<double>> log_ll;
 
  read_matrix(input_file, pp1, pp2, log_ll);
 
  shared_ptr<statistics::STR_likelihood_inputs> likelihood_inputs = static_pointer_cast<statistics::STR_likelihood_inputs>(m_likelihood_inputs);
  likelihood_inputs->interp_function2D = make_shared<glob::FuncGrid2D>(glob::FuncGrid2D(pp1, pp2, log_ll, "Spline"));
 
  m_log_likelihood_function_grid = LogLikelihood_2D_interpolated;
  m_likelihood_function_grid = [&] (vector<double> &par, const shared_ptr<void> input) { return exp(m_log_likelihood_function_grid(par, input));};
}
 
 
// ============================================================================================
 
 
cbl::statistics::Likelihood::Likelihood (const std::shared_ptr<data::Data> data, const std::shared_ptr<Model> model, const LikelihoodType likelihood_type, const std::vector<size_t> x_index, const int w_index, const std::shared_ptr<ModelParameters> model_parameters, const double prec, const int Nres)
{
  set_data(data);
  set_model(model, model_parameters);
  set_function(likelihood_type, x_index, w_index, prec, Nres);
}
 
 
// ============================================================================================
 
 
cbl::statistics::Likelihood::Likelihood (const std::shared_ptr<data::Data> data, const std::shared_ptr<Model> model, const Likelihood_function log_likelihood_function, const std::shared_ptr<ModelParameters> model_parameters)
{
  set_data(data);
  set_model(model, model_parameters);
  set_log_function(log_likelihood_function);
}
 
 
// ============================================================================================
 
 
double cbl::statistics::Likelihood::operator () (std::vector<double> &pp) const
{
  return exp(this->log(pp));
}
 
 
// ============================================================================================
 
 
double cbl::statistics::Likelihood::log (std::vector<double> &parameter) const
{
  return (m_use_grid) ? m_log_likelihood_function_grid(parameter, m_likelihood_inputs) : m_log_likelihood_function(parameter, m_likelihood_inputs);
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::set_data (const shared_ptr<data::Data> data)
{
  m_data = data;
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::set_model (const shared_ptr<Model> model, const shared_ptr<ModelParameters> model_parameters)
{
  switch (model->dimension()) {
  case Dim::_1D_:
    m_model = make_shared<Model1D>(*static_pointer_cast<Model1D>(model));
    break;
  case Dim::_2D_:
    m_model = make_shared<Model2D>(*static_pointer_cast<Model2D>(model));
    break;
  default:
    ErrorCBL("the model dimension must be Dim::_1D_ or Dim::_2D_ !", "set_model", "Likelihood.cpp");
  }
 
  m_model_parameters = (model_parameters == NULL) ?  make_shared<LikelihoodParameters>(LikelihoodParameters(m_model->parameters()->nparameters(), m_model->parameters()->type(), m_model->parameters()->name())) : model_parameters;
  m_model->set_parameters(m_model_parameters);  
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::unset_grid ()
{
  m_use_grid = false;
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::set_grid (const int npoints, const std::vector<std::vector<double>> parameter_limits, const string file, const bool read)
{
  if (m_likelihood_type==statistics::LikelihoodType::_NotSet_)
    ErrorCBL("the Likelihood function is not set!", "set_grid", "Likelihood.cpp");
 
  unsigned int npar = m_model->parameters()->nparameters_free();
  
  if (npar==0)
    ErrorCBL("there is no parameter free to vary!", "set_grid", "Likelihood.cpp");
  if (npar>2)
    ErrorCBL("wrong size for the vector of starting parameters!", "set_grid", "Likelihood.cpp");
  if (parameter_limits.size()!=npar)
    ErrorCBL("wrong size for the vector of parameter limits!", "set_grid", "Likelihood.cpp");
 
  coutCBL << "Computing tabulated likelihood!" << endl;
 
  if (npar==1) {
    if (read)
      m_set_grid_likelihood_1D(file);
    else
      m_set_grid_likelihood_1D(npoints, parameter_limits, file);
  }
  else if (npar==2) {
    if (read)
      m_set_grid_likelihood_2D(file);
    else
      m_set_grid_likelihood_2D(npoints, parameter_limits, file);
  }
 
  m_use_grid = true;
  coutCBL << "Done!" << endl;
}
 
    
// ============================================================================================
 
 
void cbl::statistics::Likelihood::set_function (const LikelihoodType likelihood_type, const std::vector<size_t> x_index, const int w_index, const double prec, const int Nres)
{
  m_x_index = x_index;
  m_w_index = w_index;
  m_likelihood_type = likelihood_type;
 
  if (m_data->dataType()==data::DataType::_1D_ || m_data->dataType()==data::DataType::_1D_extra_) {
    
    switch (m_likelihood_type)
      {
      
      case (LikelihoodType::_Gaussian_Error_):  
    for (int i=0; i<m_data->ndata(); i++) {
      if (std::isinf(pow(1./m_data->error(i), 2)))
        ErrorCBL("error("+conv(i, par::fINT)+") is too small, and 1/error("+conv(i, par::fINT)+")^2 = inf!", "set_function", "Likelihood.cpp");
      else
        continue;
    }
 
    m_log_likelihood_function = &LogLikelihood_Gaussian_1D_error;
    break; 
 
      case (LikelihoodType::_Gaussian_Covariance_):
    m_data->invert_covariance(prec, Nres);
    m_log_likelihood_function = &LogLikelihood_Gaussian_1D_covariance;
    break;
 
      case (LikelihoodType::_Poissonian_):
    m_log_likelihood_function = &LogLikelihood_Poissonian_1D_;
    break; 
 
      default:
    ErrorCBL("type of likelihood not recognized or not yet implemented!", "set_function", "Likelihood.cpp");
    break;
      }
  }
 
  else if (m_data->dataType()==data::DataType::_2D_ || m_data->dataType()==data::DataType::_2D_extra_) {
    switch (m_likelihood_type)
      {
    
      case (LikelihoodType::_Gaussian_Error_):
    for (int i=0; i<m_data->xsize(); i++) {
      for (int j=0; j<m_data->ysize(); j++) {
        if (std::isinf(pow(1./m_data->error(i, j), 2)))             
          ErrorCBL("error("+conv(i, par::fINT)+", "+conv(j, par::fINT)+") is too small, and 1/error("+conv(i, par::fINT)+", "+conv(j, par::fINT)+")^2 = inf!", "set_function", "Likelihood.cpp");
        else
          continue;
      }
    }
    m_log_likelihood_function = &LogLikelihood_Gaussian_2D_error;
    break; 
 
      case (LikelihoodType::_Poissonian_):
    m_log_likelihood_function = &LogLikelihood_Poissonian_2D_;
    break; 
 
      default:
    ErrorCBL("type of likelihood not recognized or not yet implemented!", "set_function", "Likelihood.cpp");
    break;
      }
  }
  else 
    ErrorCBL("data type not recognized or not yet implemented!", "set_function", "Likelihood.cpp");
 
  m_likelihood_function = [&] (vector<double> &par, const shared_ptr<void> input) { return exp(m_log_likelihood_function(par, input)); };
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::set_function (const Likelihood_function likelihood_function)
{
  m_likelihood_type = LikelihoodType::_UserDefined_;
  m_likelihood_function = likelihood_function;
  m_log_likelihood_function = [&] (vector<double> &par, const shared_ptr<void> input) { return std::log(m_likelihood_function(par, input)); };
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::set_log_function (const Likelihood_function log_likelihood_function)
{
  m_likelihood_type = LikelihoodType::_UserDefined_;
  m_log_likelihood_function = log_likelihood_function;
  m_likelihood_function = [&] (vector<double> &par, const shared_ptr<void> input) { return exp(m_log_likelihood_function(par, input)); };
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::maximize (const std::vector<double> start, const std::vector<std::vector<double>> parameter_limits, const unsigned int max_iter, const double tol, const double epsilon)
{
  if (m_likelihood_type==statistics::LikelihoodType::_NotSet_)
    ErrorCBL("the Likelihood function is not set!", "maximize", "Likelihood.cpp");
 
  vector<double> starting_par;
  vector<vector<double>> limits_par;
 
  unsigned int npar_free = m_model->parameters()->nparameters_free();
  unsigned int npar = m_model->parameters()->nparameters();
  
  if (start.size()==npar_free && parameter_limits.size()==npar_free) 
    {
      starting_par = start;
      limits_par = parameter_limits;
    }
  else if (start.size()==npar && parameter_limits.size()==npar)
    {
      for (size_t i=0; i<npar_free; i++) {
    starting_par.push_back(start[m_model->parameters()->free_parameter()[i]]);
    limits_par.push_back(parameter_limits[m_model->parameters()->free_parameter()[i]]);
      }
    }
  else
    ErrorCBL("check your inputs!", "maximize", "Likelihood.cpp");
 
  m_likelihood_inputs = make_shared<STR_likelihood_inputs>(STR_likelihood_inputs(m_data, m_model, m_x_index, m_w_index));
 
  function<double(vector<double> &)> ll = [this] (vector<double> &pp)
                      {
                        return -m_log_likelihood_function(pp, m_likelihood_inputs); 
                      };
 
  coutCBL << "Maximizing the likelihood..." << endl;
  vector<double> result = cbl::wrapper::gsl::GSL_minimize_nD(ll, starting_par, limits_par, max_iter, tol, epsilon);
  coutCBL << "Done!" << endl << endl;
 
  m_model_parameters->set_bestfit_values(result);
  m_model_parameters->write_bestfit_info();
  coutCBL << "log(Likelihood) = " << this->log(result) << endl << endl;
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::write_results (const string dir_output, const string file)
{
  coutCBL << "Writing results of Likelihood minimization on " << dir_output+file << endl;
  vector<double> bestFitValue = m_model_parameters->bestfit_value();
  string name = LikelihoodTypeNames ()[static_cast<int>(m_likelihood_type)];
  double likelihoodValue = this->log(bestFitValue);
 
  string mkdir = "mkdir -p "+dir_output;
  if (system(mkdir.c_str())) {}
 
  ofstream fout(dir_output+file);
 
  fout << "#Parameters information" << endl;
  fout << "nParameters = " << bestFitValue.size() << endl;
 
  for (size_t i=0; i<bestFitValue.size(); i++) {
    fout << "par" << i+1 << "_name = " << m_model_parameters->name(i) << endl;
    fout << "par" << i+1 << "_status = " << m_model_parameters->status(i) << endl;
    fout << "par" << i+1 << "_bestfit_value = " << bestFitValue[i] << endl;
  }
 
  fout << "#Likelihood information" << endl;
  fout << "likelihoodType = " << name << endl;
  fout << "logLikelihoodValue = " << likelihoodValue << endl;
 
  fout.clear(); fout.close();
  coutCBL << "I wrote the file " << dir_output+file << endl;
}
 
 
// ============================================================================================
 
 
void cbl::statistics::Likelihood::write_model (const string output_dir, const string output_file, const std::vector<double> parameters, const std::vector<double> xx, const std::vector<double> yy)
{
  switch (m_model->dimension()) {
 
  case Dim::_1D_: 
    {
      vector<double> xvec = xx;
      if (xx.size()==0)
    xvec = m_data->xx();
 
      m_model->write(output_dir, output_file, xvec, parameters);
    }
    break;
    
  case Dim::_2D_: 
    {
      vector<double> xvec = xx, yvec = yy;
      if (xx.size()==0)
    xvec = m_data->xx();
      if (yy.size()==0)
    yvec = m_data->yy();
 
      m_model->write(output_dir, output_file, xvec, yvec, parameters);
    }
    break;
    
  default:
    ErrorCBL("the input dimension must be Dim::_1D_ or Dim::_2D_!", "write_model", "Likelihood.cpp");
  }
}