PndMultiClassMlpTrain.h

Go to the documentation of this file.

//****************************************************************************
//*                   This file is part of PandaRoot.                        *
//*                                                                          *
//*            PandaRoot is distributed under the terms of the               *
//*              GNU General Public License (GPL) version 3,                 *
//*                 copied verbatim in the file "LICENSE".                   *
//*                                                                          *
//*  Copyright (C) 2006 - 2024 FAIR GmbH and copyright holders of PandaRoot  *
//*     The copyright holders are listed in the file "COPYRIGHTHOLDERS".     *
//*               The authors are listed in the file "AUTHORS".              *
//****************************************************************************

/* ***************************************
 * MultiClass MLP  Training functions    *
 * Author: M.Babai@rug.nl                *
 * Version:                              *
 * LICENSE:                              *
 * ***************************************
 */
/*
 * Note: This is just an interface to the original TMVA
 * implementation. To find out the available options, please read TMVA
 * manuals. In case of errors or wrong outputs produced by TMVA
 * classifiers, try to read their mailing list and send your questions
 * to the same list.
 ******* VERY IMORTANT ****
 * You NEED TMVA version > 4.1.X before this works.
 */
//#pragma once
#ifndef PND_MULTICLASS_MLP_TRAIN_H
#define PND_MULTICLASS_MLP_TRAIN_H

// Local includes
#include "PndMvaTrainer.h"

// TMVA && ROOT
#include "TMVA/Factory.h"
#include "TMVA/Config.h"

// Interface definition for Multiclass MLP trainers.
class PndMultiClassMlpTrain : public PndMvaTrainer {
  //----------------------------------------
  //================== public ==============
 public:
  explicit PndMultiClassMlpTrain(std::string const &InPut, std::vector<std::string> const &ClassNames, std::vector<std::string> const &VarNames, bool trim = true);
  virtual ~PndMultiClassMlpTrain();

  void Train();

  void storeWeights();

  void Initialize();

  //______________________________________________
  //====== Getters and setters.
  // Set the name of the current job
  inline void SetJobName(std::string const &name);

  // Set data Transformation scheme
  inline void SetTransformation(std::string const &tran);

  // Set the options for the MLP alg. See TMVA manuals.
  inline void SetMlpOptions(std::string const &opts);

  // Set the file name to store evaluation outputs.
  inline void SetEvalFileName(std::string const &fname);

  // Set the directory where weights are stored.
  inline void SetWeightsOutDir(std::string const &dirName);

  // Evaluate the classifier?
  inline void SetEvaluation(bool evaluate);

  // Get the current job name.
  inline std::string const &GetJobName() const;

  // Get the current transformation info.
  inline std::string const &GetTransformation() const;

  // Get the classifier options.
  inline std::string const &GetMlpOptions() const;

  // Get the name of the weight file.
  inline std::string const &GetEvalFileName() const;

  // Get the directory where the weights are stored.
  inline std::string const &GetWeightsOutDir() const;
  //----------------------------------------

  //================== protected ============
  // protected:
  //----------------------------------------

  //================== private =============
 private:
  // To avoid mistakes.
  PndMultiClassMlpTrain(PndMultiClassMlpTrain const &oth);
  PndMultiClassMlpTrain &operator=(PndMultiClassMlpTrain const &oth);

  // Initialize mlp object and set the options.
  void InitMlp();

  // Add the variables to the TMVA factory object.
  void AddVariables();

  //==============================
  TMVA::Factory *m_factory;    // TMVA factory
  TFile *EvalFile;             // To store evaluation file
  std::string m_JName;         // Job name
  std::string m_transform;     // Transformation opt.
  std::string m_MlpOptions;    // mlp options.
  std::string m_evalFileName;  // evaluation file name.
  std::string m_weightDirName; // Directory name to store weights.
  bool m_Evaluate;
}; // End of interface definition.
//=============== inline functions implementation. ========
//__________________________________________
inline void PndMultiClassMlpTrain::SetJobName(std::string const &name)
{
  this->m_JName = name;
};

inline void PndMultiClassMlpTrain::SetTransformation(std::string const &tr)
{
  this->m_transform = tr;
};

inline void PndMultiClassMlpTrain::SetMlpOptions(std::string const &opt)
{
  this->m_MlpOptions = opt;
};

inline std::string const &PndMultiClassMlpTrain::GetJobName() const
{
  return m_JName;
};

inline std::string const &PndMultiClassMlpTrain::GetTransformation() const
{
  return m_transform;
};

inline std::string const &PndMultiClassMlpTrain::GetMlpOptions() const
{
  return m_MlpOptions;
};

inline void PndMultiClassMlpTrain::SetEvalFileName(std::string const &fname)
{
  this->m_evalFileName = fname;
};

inline std::string const &PndMultiClassMlpTrain::GetEvalFileName() const
{
  return m_evalFileName;
};

inline void PndMultiClassMlpTrain::SetWeightsOutDir(std::string const &dirName)
{
  this->m_weightDirName = dirName;
};

inline std::string const &PndMultiClassMlpTrain::GetWeightsOutDir() const
{
  return m_weightDirName;
};

inline void PndMultiClassMlpTrain::SetEvaluation(bool evaluate)
{
  this->m_Evaluate = evaluate;
};
#endif