一个C++的程序怎么移植成C# - C# 论坛

得分:0

一个C++的程序怎么移植成C#

bp算法的vc++代码
#include <conio.h>
#include <math.h>
#include <stdlib.h>
#include <stdio.h>
#define Beta 1.0
#define NAME_MAX 80
double
bp(w_ih,w_ho,patt,targ,nn,hh,pp,mm,fin,fileW,
resultF,crition,total,sa)
double **w_ih;/* weight from input to hidden */
double **w_ho;/* weight from hidden to output */
double **patt;/* input patterns */
double **targ;/* desired outputs */
intnn;/* input dim */
int hh;/* hidden dim */
int pp;/* output dim */
int mm;/* training patterns number */
int*fin;/* flag for terminate */
char*fileW;/* weightFile */
char*resultF;/* resultFile Name */
doublecrition;/* minimum mean square */
int total;
int sa;/* flag for saving weights */
{
double *output;
int i,j,s,num_p,num_d,mis,mis_tr,mis_te;
double err;
char Msg[NAME_MAX],Msg1[NAME_MAX],Msg2[NAME_MAX],
Msg3[NAME_MAX],Msg4[NAME_MAX],Msg5[NAME_MAX];
FILE *fp;
FILE *cor;
extern int maximum(double *,int);
//extern void drawbp(double **,double **,int,int,int);
void feedward(double **,double **,double *,double *,int,int,int);
double errSqu(double *,double *,int);
output=(double *)calloc(pp,sizeof(double));
err=0.0;
for (i=0;i<mm;i++)
{
feedward(w_ih,w_ho,patt,output,nn,hh,pp);
err +=errSqu(output,targ,pp);
}

/* err=sqrt(err/(pp*mm)); mean-square-root */
err=err/(pp*mm); /* mean-square */
if ((err<=crition)||(sa==1))
{
/* draw bp */
//drawbp(w_ih,w_ho,nn,hh,pp);
if (err<=crition)
*fin=1;
if ((fp=fopen(fileW,"wt"))==NULL)
{
fprintf(stderr,"cannot open %s file for writing\n",fileW);
exit(1);
}
sprintf(Msg,"weight from input to hidden");
fprintf(fp,"%s\n",Msg);
for (j=0;j<hh;j++)
{
for (s=0;s<(nn+1);s++)
fprintf(fp,"%f\t",w_ih[j][s]);
fprintf(fp,"\n");
}
sprintf(Msg,"weight from hidden to output");
fprintf(fp,"%s\n",Msg);
for (j=0;j<pp;j++)
{
for (s=0;s<(hh+1);s++)
fprintf(fp,"%f\t",w_ho[j][s]);
fprintf(fp,"\n");
}
fclose(fp);
if ((cor=fopen(resultF,"wt"))==NULL)
{
fprintf(stderr,"cannot open %s for writing\n",resultF);
exit(1);
}
sprintf(Msg,"Index");
sprintf(Msg1,"Class_Get");
sprintf(Msg2,"Class_Des");
sprintf(Msg3,"Out_1");
sprintf(Msg4,"Out_2");
sprintf(Msg5,"Out_3");
fprintf(cor,"%10s %10s %10s %10s %10s %10s\n",Msg,Msg1,Msg2,Msg3,Msg4,Msg5);
mis=0;
mis_tr=0;
mis_te=0;
for (i=0;i<total;i++)
{
feedward(w_ih,w_ho,patt,output,nn,hh,pp);
num_p=maximum(output,pp);
num_d=maximum(targ,pp);
if (num_p!=num_d)
{
if (i<mm)
mis_tr++;
else
mis_te++;
mis++;
}
fprintf(cor,"%10.3d %10.1d %10.1d",i,num_p,num_d);
for (j=0;j<pp;j++)
fprintf(cor,"%10.6f",output[j]);
fprintf(cor,"\n");
}
fprintf(cor,"Total number of misclassfication: %d\n",mis);
fprintf(cor," for traing: %d\t for test: %d\n",mis_tr,mis_te);
fclose(cor);
gotoxy(57,2);
printf("Error:%6.6f",err);
gotoxy(57,3);
printf("traing:%3.3d test:%3.3d",mis_tr,mis_te);
}
free(output);
return(err);
}
void
feedward(wih,who,in,out,n1,h1,p1)
double **wih;/* weight from input to hidden */
double **who;/* weight from hidden to output */
double *in;/* input vector */
double *out;/* output vector */
intn1;/* input dim */
int h1;/* hidden dim */
int p1;/* output dim */
{
int i,j;
double sum;
double *out_h;
double sigmoid(double);

out_h=(double *)calloc(h1,sizeof(double));
/* calculate the outputs of hidden neurons */
for (i=0;i<h1;i++)
{
sum=0.0;
for (j=0;j<n1;j++)
sum +=((wih[j])*(in[j]));
sum +=wih[n1];
out_h=sigmoid(sum);
}
/* calculate the network output */
for (i=0;i<p1;i++)
{
sum =0.0;
for (j=0;j<h1;j++)
sum +=who[j]*out_h[j];
sum +=who[h1];
out=sigmoid(sum);
}
free(out_h);
}
double
errSqu(ou,ta,le)
double *ou;/* practical output */
double *ta;/* desired output */
int le;/* output dim */
{
int i;
double dif,sum;
sum=0.0;
for (i=0;i<le;i++)
{
dif=ou-ta;
sum +=(dif*dif);
}
return(sum);
}
double
sigmoid(x)
double x;
{
x=(tanh(x*Beta)+1.0)/2;
return(x); }