// LinearDiscriminantTraining.cpp: implementation of the LinearDiscriminantTraining class.
//CHEN ChangFeng's training class
//////////////////////////////////////////////////////////////////////

#include "LinearDiscriminantTraining.h"
#include "FeatureSelection.h"
#include "KNNDiscriminantTraining.h"
#include <fstream>
#include <iostream>
#include <vector>
#include <time.h> 
using namespace std;

//////////////////////////////////////////////////////////////////////
// Construction/Destruction
//////////////////////////////////////////////////////////////////////

LinearDiscriminantTraining::LinearDiscriminantTraining()
{
   int i;
   FeatureSelection f;
   f.OpenSampleFile("digit0.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup0.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup0.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w0.push_back(1);

   f.OpenSampleFile("digit1.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup1.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup1.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w1.push_back(1);

   f.OpenSampleFile("digit4.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup4.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup4.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w4.push_back(1);
 
   f.OpenSampleFile("digit5.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup5.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup5.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w5.push_back(1);

   f.OpenSampleFile("digit6.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup6.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup6.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w6.push_back(1);

   f.OpenSampleFile("digit7.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup7.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup7.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w7.push_back(1);

   f.OpenSampleFile("digit8.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup8.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup8.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w8.push_back(1);

   f.OpenSampleFile("digit9.dat");
   f.CharacVectorGeneration();
   for(i=0;i<(f.trainingGroup).size();i++)
	   trainingGroup9.push_back(f.trainingGroup[i]);
   for(i=0;i<(f.testGroup).size();i++)
	   testGroup9.push_back(f.testGroup[i]);
   for(i=0;i<f.trainingGroup[0].size();i++)
	   w9.push_back(1);

   classResult.open("RecongnizeResult.txt");

}

LinearDiscriminantTraining::~LinearDiscriminantTraining()
{

}

int LinearDiscriminantTraining::FeatureVectorProcess()
{
   int i,j;
   double h;
   for(i=0;i<trainingGroup0.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup0[i].size()-2;j++)
	   {
		   h += trainingGroup0[i][j];
	   }

	   for(j=1;j<trainingGroup0[i].size()-2;j++)
	   {
		   trainingGroup0[i][j] = trainingGroup0[i][j]/h;
	   }
   }
   for(i=0;i<trainingGroup1.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup1[i].size()-2;j++)
	   {
		   h += trainingGroup1[i][j];
	   }

	   for(j=1;j<trainingGroup1[i].size()-2;j++)
	   {
		   trainingGroup1[i][j] = trainingGroup1[i][j]/h;
	   }
   }

   for(i=0;i<trainingGroup4.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup4[i].size()-2;j++)
	   {
		   h += trainingGroup4[i][j];
	   }

	   for(j=1;j<trainingGroup4[i].size()-2;j++)
	   {
		   trainingGroup4[i][j] = trainingGroup4[i][j]/h;
	   }
   }

   for(i=0;i<trainingGroup5.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup5[i].size()-2;j++)
	   {
		   h += trainingGroup0[i][j];
	   }

	   for(j=1;j<trainingGroup5[i].size()-2;j++)
	   {
		   trainingGroup5[i][j] = trainingGroup5[i][j]/h;
	   }
   }

   for(i=0;i<trainingGroup6.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup6[i].size()-2;j++)
	   {
		   h += trainingGroup6[i][j];
	   }

	   for(j=1;j<trainingGroup6[i].size()-2;j++)
	   {
		   trainingGroup6[i][j] = trainingGroup6[i][j]/h;
	   }
   }

   for(i=0;i<trainingGroup7.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup7[i].size()-2;j++)
	   {
		   h += trainingGroup7[i][j];
	   }

	   for(j=1;j<trainingGroup7[i].size()-2;j++)
	   {
		   trainingGroup7[i][j] = trainingGroup7[i][j]/h;
	   }
   }

   for(i=0;i<trainingGroup8.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup8[i].size()-2;j++)
	   {
		   h += trainingGroup8[i][j];
	   }

	   for(j=1;j<trainingGroup8[i].size()-2;j++)
	   {
		   trainingGroup8[i][j] = trainingGroup8[i][j]/h;
	   }
   }

   for(i=0;i<trainingGroup9.size();i++)
   {
	   h =0;
	   for(j=1;j<trainingGroup9[i].size()-2;j++)
	   {
		   h += trainingGroup9[i][j];
	   }

	   for(j=1;j<trainingGroup9[i].size()-2;j++)
	   {
		   trainingGroup9[i][j] = trainingGroup9[i][j]/h;
	   }
   }
   return 0;
}

int LinearDiscriminantTraining::TrainingProcedure()
{
    int f,i;
//	FeatureVectorProcess();
    while(f<8*TRAININGNUM)
	{
	   f =0;
       f +=TrainingSingleClass(trainingGroup0,0,w0);
       f +=TrainingSingleClass(trainingGroup1,1,w1);
       f +=TrainingSingleClass(trainingGroup4,4,w4);
       f +=TrainingSingleClass(trainingGroup5,5,w5);
       f +=TrainingSingleClass(trainingGroup6,6,w6);
       f +=TrainingSingleClass(trainingGroup7,7,w7);
       f +=TrainingSingleClass(trainingGroup8,8,w8);
       f +=TrainingSingleClass(trainingGroup9,9,w9);
	}
	cout <<"Finish Training"<<endl<<endl;
	for(i=0;i<w0.size();i++)
	{
		cout<<w0[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w1.size();i++)
	{
		cout<<w1[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w4.size();i++)
	{
		cout<<w4[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w5.size();i++)
	{
		cout<<w5[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w6.size();i++)
	{
		cout<<w6[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w7.size();i++)
	{
		cout<<w7[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w8.size();i++)
	{
		cout<<w8[i]<<"  ";
	}
	cout << endl;
	for(i=0;i<w9.size();i++)
	{
		cout<<w9[i]<<"  ";
	}
	cout << endl;
	return 0;
}

int LinearDiscriminantTraining::TrainingSingleClass(vector < vector<double> >& trainingGroup
													,int class_sign,vector < double >& w)
{
   int i,j,h,g,s,t,m,
	   n = 0;
   vector< double >tmp;
   for(i=0;i<trainingGroup.size();i++)
   {
        for(j=0,h=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w[j];
			h += g;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w0[j];
			s += g;
		}
		tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w1[j];
			s += g;
		}
        tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w4[j];
			s += g;
		}
        tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w5[j];
			s += g;
		}
        tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w6[j];
			s += g;
		}
        tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w7[j];
			s += g;
		}
        tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w8[j];
			s += g;
		}
        tmp.push_back(s);
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w9[j];
			s += g;
		}
        tmp.push_back(s);
		m = 0;
		for(t=0;t <tmp.size();t++)
		{
		    if(h > tmp[t])
			 m++;
		}
		if(m <tmp.size()-1)
		{
		   for(j=0;j<w.size();j++)
			   w[j] += trainingGroup[i][j]*STEP_LENGTH;
		}
		else n++;
		tmp.clear();
   }
   return n;
}

double LinearDiscriminantTraining::TestingProcedure()
{
	double h = 0;
    double s;
    time_t begin, end;
    float TrainingTime =0, TestingTime =0;
    begin=time(NULL);  ///
	TrainingProcedure();
    end=time(NULL);  ///
    TrainingTime= difftime(begin,end); ///

	begin=time(NULL);
    s =TestingSingleClass(testGroup0,0);
	end=time(NULL);
	TestingTime += difftime(begin,end); ///
    h +=s;
	s = s/TESTNUM;
	classResult <<"The reconginize precision of 0:"<<s<<" "<<endl;

	begin =time(NULL);
	s =TestingSingleClass(testGroup1,1);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
    h +=s;
	s = s/TESTNUM;
	classResult <<"The reconginize precision of 1:"<<s<<" "<<endl;

	begin =time(NULL);
	s =TestingSingleClass(testGroup4,4);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
    h +=s;
	s = s/TESTNUM;
	classResult <<"The reconginize precision of 4:"<<s<<" "<<endl;

	begin =time(NULL);
    s =TestingSingleClass(testGroup5,5);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
	h +=s;
	s =s/TESTNUM;
	classResult <<"The reconginize precision of 5:"<<s<<" "<<endl;

	begin =time(NULL);
    s=TestingSingleClass(testGroup6,6);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
	h +=s;
	s =s/TESTNUM;
	classResult <<"The reconginize precision of 6:"<<s<<" "<<endl;

	begin =time(NULL);
    s =TestingSingleClass(testGroup7,7);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
    h +=s;
	s =s/TESTNUM;
	classResult <<"The reconginize precision of 7:"<<s<<" "<<endl;

	begin =time(NULL);
    s =TestingSingleClass(testGroup8,8);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
    h +=s;
	s =s/TESTNUM;
	classResult <<"The reconginize precision of 8:"<<s<<" "<<endl;

	begin =time(NULL);
    s=TestingSingleClass(testGroup9,9);
	end =time(NULL);
	TestingTime += difftime(begin,end); ///
    h +=s;
	s =s/TESTNUM;
	classResult <<"The reconginize precision of 9:"<<s<<" "<<endl;
	classResult<<endl;
	classResult <<"The total recongnize precision: ";
	classResult << h/(8*TESTNUM)<<endl;
	classResult <<"The training time is: "<<TrainingTime<<endl;
	classResult <<"The test time is: "<<TestingTime<<endl;
	classResult.close();
	return h/(8*TESTNUM);
}

int LinearDiscriminantTraining::TestingSingleClass(vector < vector<double> >& trainingGroup
													,int class_sign)
{
   int  i,j,m=0,s,h,g,get_figure = 0;
   for(i=0;i<trainingGroup.size();i++)
   {
	    h = 0;
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w0[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 0;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w1[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 1;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w4[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 4;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w5[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 5;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w6[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 6;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w7[j];
			s += g;
		}
         if(h <= s)
		{
		    h = s;
		    get_figure  = 7;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w8[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 8;
		}
        for(j=0,s=0;j<trainingGroup[i].size();j++)
		{
		    g = trainingGroup[i][j]*w9[j];
			s += g;
		}
        if(h <= s)
		{
		    h = s;
		    get_figure  = 9;
		}
        classResult <<get_figure<<"  ";
		if(get_figure == class_sign)
			m++;
   }
   classResult <<endl;
   return m;
}

int main()
{
   int i,j;
   FeatureSelection s;
   ofstream out;
   out.open("out.txt");
   s.OpenSampleFile("digit0.dat");
   s.CharacVectorGeneration();
   for(i=0;i<(s.trainingGroup).size();i++)
   {
	   cout << i+1 << "  ";
       for(j=0;j<(s.trainingGroup)[i].size();j++)
	   {
	       cout << (s.trainingGroup)[i][j] <<"  ";
	   }
	   cout <<endl;
   }
   for(i=0;i<(s.testGroup).size();i++)
   {
	   cout << i+31 << "  ";
       for(j=0;j<(s.testGroup)[i].size();j++)
	   {
	       cout << s.testGroup[i][j] <<"  ";
	   }
	   cout <<endl;
   }
   cout <<endl <<endl;
   KNNDiscriminantTraining l;
   l.TestingProcedure();
   cout <<"Finish all..."<<endl;
   getchar();
   return 0;
}