XSM_MLP.hxx

#ifndef _XSM_MLP_HXX
#define _XSM_MLP_HXX

/*!
 \file
 \brief Header file for XSM_MLP class.
 
 
 @authors BaM
 @authors BLG
 @version 1.0
 */
#include "XSModel.hxx"
#include "CLASSReader.hxx"
#include "TTree.h"
#include <string>
#include <fstream>
#include <iostream>
#include <map>
#include <vector>


typedef long long int cSecond;
using namespace std;


class XSM_MLP;
#ifndef __ROOTCLING__
typedef void (XSM_MLP::*XS_MLP_DMthPtr)( const string & ) ;
#endif
//-----------------------------------------------------------------------------//
//! Defines a XSModel getting mean cross sections from neural network execution

/*!
 Define a XSM_MLP.
 This is the class to predict cross sections with a
 set of Multi Layer Perceptrons (MLP)
 
 @authors BLG
 @version 1.0
 */
//________________________________________________________________________


class XSM_MLP : public XSModel
{
	public :
	
	/*!
	 \name Constructor/Desctructor
	 */
	//@{
	XSM_MLP() {};
	
	//{
	/// Normal Constructor
	/*!
	 \param TMVA_Weight_Directory : The directory where all the TMVA weight are located
	 \param InformationFile : Name of the information file located in TMVA_Weight_Directory (default : Data_Base_Info.nfo)
	 
	 */
	XSM_MLP(string TMVA_Weight_Directory,string InformationFile = "/Data_Base_Info.nfo");
	//}
	
	//{
	/// CLASSLogger Constructor
	/*!
	 \param log : The CLASSLogger
	 \param TMVA_Weight_Directory : The directory where all the TMVA weight are located
	 \param InformationFile : Name of the information file located in TMVA_Weight_Directory (default : Data_Base_Info.nfo)
	 
	 */
	XSM_MLP(CLASSLogger* Log,string TMVA_Weight_Directory,string InformationFile = "/Data_Base_Info.nfo");
	//}
	
	~XSM_MLP();
	//@}
	
	/*!
	 \name Reading NFO related Method
	 */
	//@{
	
	//{
	/// LoadKeyword() : make the correspondance between keyword and reading method
	virtual void LoadKeyword();
	//}
	
	//{
	/// ReadTimeSteps : read the time step of the model
	/*!
	 \param line : line suppossed to contain the time step information starts with "k_timestep" keyword
	 */
	void ReadTimeSteps(const string &line);
	//}
	
	//{
	/// ReadZAIName : read the zai name in the TMWA MLP model
	/*!
	 \param line : line suppossed to contain the ZAI name  starts with "k_zainame" keyword
	 */
	void ReadZAIName(const string &line);
	//}
	//{
	/// ReadLine : read a line
	/*!
	 \param line : line to read
	 */
	void ReadLine(string line);
	//}
	
	//@}
	
	
  virtual void BookTMVAReader(); //Book TMVA method  
	
  EvolutionData GetCrossSections(IsotopicVector IV, double t=0);	//!< Return calculated cross section by the MLP regression
	
	
	protected :
	
  vector<CLASSReader*> fReader;
	
  void GetMLPWeightFiles();				//!< Find all .xml file in TMVA_Weight_Directory
	
	void ReadWeightFile(string Filename, int &Z, int &A, int &I, int &Reaction) ;				//!<  Select the reaction according to the weight file name
	
	
	
	double ExecuteTMVA(string WeightFile, TTree* InputTree);			//!<Execute the MLP according to the input tree created
	virtual vector<float> CreateTMVAInput(IsotopicVector isotopicvector, int t);	//!<Create input tmva tree to be read by ExecuteTMVA
	
	
	vector<double> 	fMLP_Time;	//!<  Time vector of the data base
	vector<string> 	fWeightFiles;	//!<  All the weight file contains in fTMVAWeightFolder
	
	string fTMVAWeightFolder;	//!<  folder containing all the weight file
	
	
	bool fIsStepTime;		//!<  true if one TMVA weihgt per step time is requiered otherwise it assumes time is part of the MLP inputs
	
	map<ZAI,string> fMapOfTMVAVariableNames;//!<  List of TMVA input variable names (read from fMLPInformationFile ) , name depends on the training step
	
#ifndef __ROOTCLING__
	map<string, XS_MLP_DMthPtr> fDKeyword;
#endif
};

#endif