NeuralNetworks_Weather

main.cpp

#include <cstdio>
#include <cstdlib>
#include <cmath>
#include <ctime>
#include <iostream>
#include <unistd.h>
using namespace std;
#define NUM_LEARN 	80000000
#define NUM_SAMPLE 	310
#define NUM_INPUT 	3
#define NUM_HIDDEN 8
#define NUM_OUTPUT	1
#define RTMAX 		50


double sigmoid(double x)
{
	return 1.0 / (1.0+exp(-x));
}

struct NN
{
	double x[NUM_INPUT];
	double y[NUM_OUTPUT];
};

unsigned int w = 123426572;
unsigned int xor128() 
{
	static unsigned int x = 123456789, y = 362436069, z = 521288629;
	unsigned int t = (x ^ (x << 11) );
	x=y; y=z; z=w;
	return ( w=(w^(w>>19)) ^ ( t ^ ( t >>8)) );
}

NN txy[NUM_SAMPLE] = { 0 };
double x[NUM_INPUT+1],h[NUM_HIDDEN+1],y[NUM_OUTPUT];
double w1[NUM_INPUT+1][NUM_HIDDEN],w2[NUM_HIDDEN+1][NUM_OUTPUT];
double h_back[NUM_HIDDEN+1],y_back[NUM_OUTPUT];

int main(int argc, char *argv[])
{
	//if(argc < 2) return -1;
	FILE *fp = fopen("data.csv", "r");
 	int learn, sample, i = 0, j, k = 0;
 	double net_input, starttm, endtm;
 	double emax;
 	double EPSILON = 0.006;
 	cin >> emax;
 	
	srand((unsigned)time(NULL));

//	goto Break;

	fscanf(fp, "%lf %lf %lf %lf", 
			&txy[0].x[0], &txy[0].x[1], &txy[0].x[2], 
			 &txy[0].y[0] );
	txy[0].x[0] = (fabs(txy[0].x[0]) ) / RTMAX;
	txy[0].x[1] = (fabs(txy[0].x[1]) ) / RTMAX;
	txy[0].x[2] = (fabs(txy[0].x[2]) ) / RTMAX;
	txy[0].y[0] = (fabs(txy[0].y[0]) ) / RTMAX;

	for(i = 1; i < NUM_SAMPLE-1; i++ )
	{
		txy[i].x[0] = txy[i-1].x[1];
		txy[i].x[1] = txy[i-1].x[2];
		txy[i].x[2] = txy[i-1].y[0];
		fscanf(fp, "%lf ", &txy[i].y[0] );
		txy[i].y[0] = (fabs(txy[i].y[0]) ) / RTMAX;
		cout << txy[i].x[0] * RTMAX << endl;
	}
	getchar(); getchar();

	for(i = 0; i < NUM_INPUT+1; i++)
		for(j = 0;j < NUM_HIDDEN; j++)
		{
			w1[i][j]= (double) ( (double) xor128() / 0xFFFFFFFF - 0.5) / 10.0;
			cout << xor128() << endl;
		}

	for(i = 0; i < NUM_HIDDEN+1; i++)
		for(j = 0;j < NUM_OUTPUT; j++)
			w2[i][j]= (double) ( (double) xor128() / 0xFFFFFFFF - 0.5) / 10.0;

	starttm = clock();
	for(learn = 0; learn < NUM_LEARN; learn++)
	{
		double e = 0.0;

		for(sample = 0; sample < NUM_SAMPLE; sample++)
		{
			for(i = 0;i < NUM_INPUT; i++)
				x[i] = txy[sample].x[i];
	
			x[NUM_INPUT] = (double)1.0;
	
			for( j = 0; j < NUM_HIDDEN;j++)
			{
			 	net_input = 0;
			 	for( i = 0; i < NUM_INPUT+1; i++ )
			 		net_input = net_input + w1[i][j] * x[i];

			 	h[j] = tanh(net_input);
     			}
	
			h[NUM_HIDDEN] = (double)1.0;
			
			for (j = 0; j < NUM_OUTPUT;j++)
			{
      			net_input = 0;
     				for(i = 0; i < NUM_HIDDEN + 1; i++)
					net_input = net_input + w2[i][j] * h[i];
				
				y[j] = sigmoid(net_input);
				e += 0.5 * (txy[sample].y[j] - y[j]) * (txy[sample].y[j] - y[j]);

				if(learn % 10000 == 0)
				{
					printf("> %lf %lf error=%lf Epsilon=%lf\n", y[j] * RTMAX, fabs(txy[sample].y[j] - y[j]), 
						e, EPSILON);
				} 
			}
						
  			for(i = 0; i < NUM_HIDDEN;i++)
			{
				net_input = 0;
				
				for(j = 0; j < NUM_OUTPUT; j++)
				{
					y_back[j] = ( y[j] - txy[sample].y[j] ) * ( 1.0 - y[j] ) * y[j];
					net_input = net_input + w2[i][j] * y_back[j];
				}
	
				h_back[i] = net_input * ((double)1.0 - h[i]) * h[i];
			}
		
  			for(i = 0; i < NUM_INPUT+1; i++)
  				for(j = 0; j < NUM_HIDDEN; j++)
					w1[i][j] = w1[i][j] - EPSILON * x[i] * h_back[j];
  			for(i = 0; i < NUM_HIDDEN+1; i++)
  				for(j = 0; j < NUM_OUTPUT; j++)
    					w2[i][j] = w2[i][j] - EPSILON * h[i] * y_back[j];
    		}
    	//	if(learn % 10000 == 0)
    		/*if(e - 0.01 <  emax) 
    			EPSILON = e - emax + 0.001;*/

    		if(e <= emax)
    		{
    			cout << "error = " << e << endl;
    			break;
    		}
	}

Break:

	fclose(fp);
	fseek(fp, 0L, SEEK_SET);
	fp = fopen("w_io.wd", "wb");
	if(!fp) cout << "err" << endl;
		fwrite(w1, sizeof(double) , (NUM_INPUT+1) * NUM_HIDDEN, fp);
		fwrite(w2, sizeof(double) , (NUM_HIDDEN+1) * NUM_OUTPUT, fp);
	fclose(fp);

	endtm = clock();
	cout << "End of learning : " << (endtm-starttm) / CLOCKS_PER_SEC / 60 << "minutes" << endl;

	for(int w = 0; w < 10; w++)
	{
		for(i = 0;i < NUM_INPUT; i++)
		{
			cin >> x[i];
			x[i] /= RTMAX;
		}
		for(k=0; k<30; k++)
		{	
			h[NUM_HIDDEN] = x[NUM_INPUT] = (double)1.0;
	
			for( j = 0; j < NUM_HIDDEN;j++)
			{
			 	net_input = 0;
			 	for(i = 0; i < NUM_INPUT+1; i++)
			 	{
			 		net_input = net_input + w1[i][j] * x[i]; 
			 	}
			 	h[j] = tanh(net_input);
     			}
		
			for (j = 0; j < NUM_OUTPUT;j++)
			{
     		 		net_input = 0;
		
     				for(i = 0;i < NUM_HIDDEN+1; i++)
					net_input += w2[i][j] * h[i];
		
				y[j] = sigmoid(net_input);
				printf("> %lf\n", y[j] * RTMAX);
			}
			for (j = 0; j < NUM_INPUT-1; j++)
			{
				x[j] = x[j+1];
			}
			x[NUM_INPUT-1] = y[0];
		}
	}

	return 0;
}

/*
0.3.2
0.3.4
0.3.7
0.3.8
0.3.5
0.3.8
0.3.6
0.3.9
0.3.8
0.3.4
0.3.7
0.3.3
0.3.2
0.2.9
0.2.8
0.2.9
0.2.6
0.2.8
0.2.5
0.2.7
0.1.9
0.2.4
0.2.4
0.2.2
0.2.3
0.2.1
0.2.1
0.2.1
0.2.3
0.2.3
0.2.3
0.2.4
0.2.1
0.2.5
0.2.7
0.2.9
0.3.2
0.3.0
0.3.0
0.3.0
0.3.1
0.3.0
0.3.1
0.3.1
0.3.0
0.3.1
0.3.2
0.3.3
0.3.2
0.3.4
0.3.5
0.3.6
0.4.0
0.4.4
0.5.1
0.5.4
0.6.0
0.5.8
0.6.6
0.7.6
0.8.0
0.8.0
0.7.7
0.7.9
0.19.2
0.9.5
0.18.1
0.18.7
0.8.7
0.9.7
0.19.7
0.110.
0.110.
0.10.
0.11.
0.10.
0.10.
0.110.
0.11.
0.12.
0.11.
0.11.
0.11.
0.11.
0.10.
0.10.
0.97
0.96
0.95
0.95
0.96
0.91
0.87
0.85
0.84
0.83
0.82
0.81
0.78
0.77
0.75
0.72
0.70
0.73
0.77
0.79
0.77
0.76
0.76
0.77
0.73
0.70
0.68
0.70
0.73
0.70
0.68
0.72
0.69
0.70
0.72
0.68
0.67
0.69
0.72
0.65
0.67
0.70
0.68
0.66
0.64
0.66
0.65
0.64
0.63
0.61
0.65
0.63
0.56
0.58
0.55
0.51
0.53

*/