/****************************************************************************************/
/*                                                                                      */
/*  Project     :  1D-BRNN                                                              */
/*  Release     :  3.3                                                                  */
/*                                                                                      */
/*  File        :  Dataset.cpp                                                          */
/*  Description :  Dataset of sequences in input of the program                         */
/*                                                                                      */
/*  Author(s)   :  Christophe Magnan  (2013) - New generic version                      */
/*              :  Jianlin Cheng      (2003) - New custom version for SCRATCH           */
/*              :  Gianluca Pollastri (2001) - Customized version for SCRATCH           */
/*              :  Paolo Frasconi     (1997) - Initial generic version                  */
/*                                                                                      */
/*  Copyright   :  Institute for Genomics and Bioinformatics                            */
/*                 University of California, Irvine                                     */
/*                                                                                      */
/*  Modified    :  2015/07/01                                                           */
/*                                                                                      */
/****************************************************************************************/

#include "Dataset.h"

/*******************************************/
/*                Interface                */
/*******************************************/

// Constructor
Dataset::Dataset() :
  Num_Sequences(-1),Num_Positions(0),Num_Features(-1),
  Num_Classes(-1),Sequences(NULL){}

// Destructor
Dataset::~Dataset()
{
  for(int i=0;i<Num_Sequences;i++){
  delete Sequences[i]; }
  delete[] Sequences;
}

// Load dataset
void Dataset::load(char* filein)
{
  filebuf in_buffer;
  if(in_buffer.open(filein,ios::in)!=0)
  { istream is(&in_buffer);
    is >> Num_Sequences; is >> Num_Features;
    is >> Num_Classes; Num_Positions=0;
    Sequences=new Sequence*[Num_Sequences];
    for(int i=0;i<Num_Sequences;i++)
    { Sequences[i]=new Sequence();
      Sequences[i]->load(is,Num_Features,Num_Classes);
      Num_Positions+=Sequences[i]->Num_Positions; }
    in_buffer.close(); check_viability(); }
  else{ cerr << "Cannot read file \"" << filein << "\"\n"; exit(1); }
}

// Allocate memory for propagation
void Dataset::alloc_propagation(int out_fwd,int out_bwd)
{
  for(int i=0;i<Num_Sequences;i++){
  Sequences[i]->alloc_propagation(out_fwd,out_bwd); }
}

// Allocate memory for back-propagation
void Dataset::alloc_backpropagation(int out_fwd,int out_bwd)
{
  for(int i=0;i<Num_Sequences;i++){
  Sequences[i]->alloc_backpropagation(out_fwd,out_bwd); }
}

// Write the model predictions
void Dataset::write_predictions(char* fileout)
{
  filebuf out_buffer; if(out_buffer.open(fileout,ios::out)!=0){
  ostream os(&out_buffer); for(int i=0;i<Num_Sequences;i++){
  Sequences[i]->write_predictions(os); } out_buffer.close(); }
  else{ cerr << "Cannot write into file \"" << fileout << "\"\n"; exit(1); }
}

// Display the confusion matrix for the last predictions
void Dataset::display_confusion()
{
  int i,j,cor=0,n=Num_Classes; float acc; int** conf; conf=new int*[n];
  for(i=0;i<n;i++){ conf[i]=new int[n]; for(j=0;j<n;j++){ conf[i][j]=0; } }
  for(i=0;i<Num_Sequences;i++){ Sequence* s=Sequences[i];
  for(j=0;j<s->Num_Positions;j++){ conf[s->Classes[j]][s->Predictions[j]]++; } }
  for(i=0;i<n;i++){ cor+=conf[i][i]; } acc=(float)cor/Num_Positions*100.0;
  cerr << cor << "/" << Num_Positions << " correctly classified";
  cerr << " (" << acc << "%)\n"; for(i=0;i<n;i++){ cerr << "  ";
  for(j=0;j<n;j++){ cerr << " " << conf[i][j]; } cerr << "\n"; } cerr << "\n";
}

// Compute the empirical risk for the last predictions
float Dataset::empirical_risk()
{
  float error=0.0; for(int i=0;i<Num_Sequences;i++){
  Sequence* s=Sequences[i]; for(int p=0;p<s->Num_Positions;p++){
  error+=s->prediction_error(p); } } return error;
}

// Check the viability of the dataset
void Dataset::check_viability()
{
  if((Num_Sequences<=0)||(Num_Positions<=0)||(Sequences==NULL)){
  cerr << "\n\nERROR : the input dataset is empty\n"; exit(1); }
}