码迷,mamicode.com
首页 > 编程语言 > 详细

bp神经网络的实现C++

时间:2017-05-20 23:38:19      阅读:262      评论:0      收藏:0      [点我收藏+]

标签:cout   ++   div   c++   ror   double   code   bp神经网络   val   

#include <iostream>
#include<stdlib.h>
#include <math.h>
using namespace std;

#define  innode 2  
#define  hiddennode 10
#define  outnode 1 
#define  sample 4
class bpnet
{
public:
    double w1[hiddennode][innode];
    double w2[outnode][hiddennode];
    double b1[hiddennode];
    double b2[outnode];
    double e;
    double error;
    double lr;
    bpnet();
    ~bpnet();
    void init();
    double randval(double low, double high);
    void initw(double w[], int n);
    void train(double p[sample][innode], double t[sample][outnode]);
    double sigmod(double y);
    double dsigmod(double y);
    void predict(double p[]);
};
double bpnet::dsigmod(double y)
{
    return y*(1 - y);
}
double bpnet::sigmod(double y)
{
    return 1.0 / (1 + exp(-y));
}
double bpnet::randval(double low, double high)
{
    double val;
    val = ((double)rand() / (double)RAND_MAX)*(high - low) + low;
    return val;
}
void bpnet::initw(double w[], int n)
{
    for (int i = 0; i < n; i++)
    {
        w[i] = randval(-0.01, 0.01);
    }
}
void bpnet::init()
{
    initw((double*)w1, hiddennode*innode);
    initw((double*)w2, hiddennode*outnode);
    initw(b1, hiddennode);
    initw(b2, outnode);
}
void bpnet::train(double p[sample][innode], double t[sample][outnode])
{
    double hiddenerr[hiddennode];
    double outerr[outnode];
    double hiddenin[hiddennode];
    double hiddenout[hiddennode];
    double outin[outnode];
    double outout[outnode];
    double x[innode];
    double d[outnode];
    for (int k = 0; k < sample; k++)
    {
        for (int i = 0; i < innode; i++)
        {
            x[i] = p[k][i];
        }
        for (int i = 0; i < outnode; i++)
        {
            d[i] = t[k][i];
        }
        for (int i = 0; i < hiddennode; i++)
        {
            hiddenin[i] = 0.0;
            for (int j = 0; j < innode; j++)
            {
                hiddenin[i] += w1[i][j] * x[j];
            }
            hiddenout[i] = sigmod(hiddenin[i] + b1[i]);
        }
        for (int i = 0; i < outnode; i++)
        {
            outin[i] = 0.0;
            for (int j = 0; j < hiddennode; j++)
            {
                outin[i] += w2[i][j] * hiddenout[j];
            }
            outout[i] = sigmod(outin[i] + b2[i]);
        }
        for (int i = 0; i < outnode; i++)
        {
            outerr[i] = (d[i] - outout[i])*dsigmod(outout[i]);
            for (int j = 0; j < hiddennode; j++)
            {
                w2[i][j] += lr*outerr[i] * hiddenout[j];
            }
        }
        for (int i = 0; i < hiddennode; i++)
        {
            hiddenerr[i] = 0.0;
            for (int j = 0; j < outnode; j++)
            {
                hiddenerr[i] += w2[j][i] * outerr[j];
            }
            hiddenerr[i] = hiddenerr[i] * dsigmod(hiddenout[i]);
            for (int j = 0; j < innode; j++)
            {
                w1[i][j] += lr*hiddenerr[i] * x[j];
            }
        }
        for (int i = 0; i < outnode; i++)
        {
            e += pow((d[i] - outout[i]), 2);
        }
        error = e / 2.0;
        for (int i = 0; i < outnode; i++)
        {
            b2[i]=lr*outerr[i];
        }
        for (int i = 0; i < hiddennode; i++)
        {
            b1[i] =hiddenerr[i] * lr;
        }
    }
}
void bpnet::predict(double p[])
{
    double hiddenin[hiddennode];
    double hiddenout[hiddennode];
    double outin[outnode];
    double outout[outnode];
    double x[innode];
    for (int i = 0; i < innode; i++)
    {
        x[i] = p[i];
    }
    for (int i = 0; i < hiddennode; i++)
    {
        hiddenin[i] = 0.0;
        for (int j = 0; j < innode; j++)
        {
            hiddenin[i] += w1[i][j] * x[j];
        }
        hiddenout[i] = sigmod(hiddenin[i] + b1[i]);
    }
    for (int i = 0; i < outnode; i++)
    {
        outin[i] = 0.0;
        for (int j = 0; j < hiddennode; j++)
        {
            outin[i] += w2[i][j] * hiddenout[j];
        }
        outout[i] = sigmod(outin[i] + b2[i]);
    }
    for (int i = 0; i < outnode; i++)
    {
        cout << "the prediction is"<<outout[i] << endl;
    }
}
bpnet::bpnet()
{
    e = 0.0;
    error = 1.0;
    lr = 0.4;
}
bpnet::~bpnet()
{}

double X[sample][innode] = {
    {1,1},
    {1,0},
    {0,1},
    {0,0}
};
double Y[sample][outnode] = {
    {1},
    {0},
    {0},
    {1}
};
int main()
{
    bpnet bp;
    bp.init();
    int times = 0;
    while (bp.error > 0.001&&times <10000)
    {
        bp.e = 0.0;
        times++;
        bp.train(X, Y);
    }
    double m[2] = { 0,1 };
    bp.predict(m);
    return 0;
}

 

bp神经网络的实现C++

标签:cout   ++   div   c++   ror   double   code   bp神经网络   val   

原文地址:http://www.cnblogs.com/semen/p/6883438.html

(0)
(0)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!