我正在测试数字识别的编码。这是我的源代码。 P / s：我编辑了我的编码。这是我主要的完整编码。我尝试为sigmoid函数添加public / protected / private，但错误不断增加。
import java.util.Scanner;
import java.io.*;
import java.util.*;
import javax.swing.JOptionPane;
public class Recog
{
public static void main(String[] args)
{
int row, col, data;
int[][] input = new int[10][35]; //array for input data
int[][] target = new int[10][10]; //array for target data
double[][] weight1 = new double[20][35]; //weight between input & hidden layer
double[][] weight2 = new double[10][20]; //weight bween hidden & output layer
double[] threshold1 = new double[20];
double[] threshold2 = new double[10]; //array for threshold value
double[] error = new double[20]; //error
double[] errorgradient1 = new double[20]; //error gradient between input & hidden layer
double[] errorgradient2 = new double[10]; //error gradient between hidden & output layer
double alpha=0.9;
double randomNumber = Math.random();
double randomMax = Math.random();
System.out.println("---------------------------------------------------------");
System.out.println(" Initialize weight & threshold at hidden layer ");
System.out.println("---------------------------------------------------------");
for(row=0; row<20; row++)
{
System.out.println("Initialization of weighted values at neuron.");
for(col=0; col<35; col++)
{
weight1[row][col]=((randomNumber/randomMax)*2.4)/35;
if((randomNumber%2+1)==1) //if 1 becomes negative
{
weight1[row][col]=weight1[row][col]*(-1);
}
System.out.println("Value of Hidden layer - neuron W " +row +" " +col +"[" +weight1[row][col] +"]");
}
threshold1[row]=(randomNumber/randomMax)*0.5;
if ((randomNumber%3+1)==1)
{
threshold1[row]=threshold1[row]*(-1);
System.out.println("Initialization of threshold values of neuron :");
System.out.println("initialization of neuron value : " +row +" [" +threshold1[row] +"]");
}
System.out.println("End of Neuron (1)");
//System.in.read();
}
System.out.println("---------------------------------------------------------");
System.out.println(" Initialize weight & threshold at output layer ");
System.out.println("---------------------------------------------------------");
for(row=0;row<10;row++)
{
System.out.println("Initialization of weighted values at neuron");
for(col=0;col<20;col++)
{
weight2[row][col]=((randomNumber/randomMax)*2.4)/35;
if((randomNumber%2+1)==1) //if 1 then negative
{
weight2[row][col]=weight2[row][col]*(-1);
}
System.out.println("Value of Output Layer - neuron W " +row +" " +col +" [" +weight2[row][col] +"]");
}
threshold2[row]=((randomNumber/randomMax) * 0.5);
if((randomNumber%3+1)==1) //if 1 then negative
{
threshold2[row]=threshold2[row]*(-1);
System.out.println("Initialization of threshold values at neuron : " +row);
System.out.println("threshold value at neuron : " +threshold2[row]);
}
System.out.println("End of Neuron (2)");
String fileName="number.txt"; //Name of the file
try
{
FileReader inFile = new FileReader(fileName);
BufferedReader bufferReader = new BufferedReader(inFile);
String line;
while ((line = bufferReader.readLine()) != null) // Read file line by line and print on the console
{
for(row=0; row<10; row++)
{
for (col=0; col<35; col++)
{
System.out.println(input[row][col] +" ");
}
System.out.println("Row : " +row);
}
}
bufferReader.close(); //Close the buffer reader
}
catch(Exception x) //if cannot read file
{
System.out.println("Error while reading file line by line:" + x.getMessage());
}
String fileName2=("target.txt"); //read target file
try
{
FileReader inFile2 = new FileReader(fileName2);
BufferedReader bufferReader = new BufferedReader(inFile2);
String line2;
while ((line2 = bufferReader.readLine()) != null) // Read file line by line and print on the console
{
for(row=0; row<10; row++)
{
for (col=0; col<10; col++)
{
System.out.println(target[row][col] +" ");
}
}
}
bufferReader.close(); //Close the buffer reader
}
catch(Exception x) //if cannot read file
{
System.out.println("Error while reading file line by line:" + x.getMessage());
}
//iteration-------------------------------------------------------------------------------
int epoch=0;
int milestone=1000;
while (epoch<1000000)
{
for(data=0; data<10; data++){} // end data
//learning process
System.out.println("----------LEARNING PROCESS STARTS HERE----------");
double[] activation_hidden = new double[20];
double[] activation_output = new double[10];
double temp_dotproduct=0;
double[][] deltaweight1 = new double[10][20];
double[][] deltaweight2 = new double[20][35];
double dot;
int neuron=0;
data=0;
//start
int epoch=0;
int milestone=1000;
while (epoch<1000000)
{
for (data=0; data<10; data++)
{
//test activation for all data
for (data=0; data<20; data++) //close at the end of network output
{
//test for first data
for (row=0; row<20; row++)
{
//do summation weight * input
for (col=0; col<35; col++)
{
dot=weight1[row][col] * input[data][col];
temp_dotproduct = temp_dotproduct + dot;
}
//activate the neuron when dot product of input x weight is finished
activation_hidden[row] = sigmoid(temp_dotproduct-threshold1[row]);
//reinitialize temp for the next neuron activation
temp_product=0;
}
for(row=0; row<10; row++)
{
for(col=0; col<20; col++)
{
dot = activation_hidden[col] * weight2[row][col];
temp_dotproduct = temp_dotproduct+dot;
}
activation_output[row] = sigmoid(temp_dotproduct-threshold2[row]);
//reinitialize temp for the next neuron activation
temp_dotproduct=0;
}
//error is calculated by ---> error = desired-actual <---
double errortemp=0;
//calculate error of each output neuron
// REMEMBER ! each neuron has their own error value
for(row=0; row<10; row++)
{
error[row]=target[data][row] - activation_output[row];
errortemp = error[row];
System.out.println("Error at neuron " +row +" is " +errortemp);
}
//next process is weight update - need to calculate the error gradient first and the network error(d-a)
//calculating error gradient
for (row=0; row<10; row++)
{
errorgradient2[row] = activation_output[row] * (1 - acivation_output[row]) * error[row];
errortemp = errorgradient1[row];
System.out.println("Error gradient at output neuron " +row +" is " +errortemp[row]);
}
//calculating error gradient first
for (row=0; row<10; row++)
{
errorgradient2[row] = activation_output[row] * (1 - acivation_output[row]) * error[row];
errortemp = errorgradient1[row];
}
//calculating weight corrections
//dw[outputneuron][hiddenneuron]
for (col=0; col<10; col++)
{
for (row=0; row<20; row++)
{
deltaweight1[col][row] = alpha * activation_hidden[row] * errorgradient1[col];
}
}
//calculate error gradient at hidden layer
int row1;
for (row1=0; row1<20; row++)
{
//calculate the hidden first
double sumOfErrorGradientTimesWeightOutput = 0;
for (col=0; col<10; col++)
{
for(row=0; row<20; row++)
{
sumOfErrorGradientTimesWeightOutput = errorgradient1[col] * weight2[col][row];
}
}
errorgradient2[20] = activation_hidden[row] * (1-activation_hidden[row]) * sumOfErrorGradientTimesWeightOutput;
}
//calculating weight corrections
//input[samplesize][inputneuron]
//delta[hiddenneuron][inputneuron]
for (col=0; col<20; col++)
{
for (row=0; row<35; row++)
{
deltaweight2[col][row] = alpha * input[data][row] * errorgradient1[col];
}
}
//update the weights
for (row=0; row<20; row++)
{
for (col=0; col<35; col++)
{
weight2[row][col] = weight2[row][col] + deltaweight2[row][col];
}
} //hidden weight
for (row=0; row<20; row++)
{
for (col=0; col<35; col++)
{
weight2[row][col] = weight2[row][col] + deltaweight2[row][col];
}
} //output weight
System.out.println("Epoch : " +epoch);
//end of learning process
epoch++;
if (epoch==milestone)
{
System.out.println(epoch);
milestone = milestone + 1000;
}
} //end epoch
System.out.println("---------------------------------------------------------");
System.out.println(" Testing the Input Samples ");
System.out.println("---------------------------------------------------------");
System.out.println("Enter Value between 0 - 9");
answer = Integer.parseInt();
if (data<=10 && data>=0)
{
for (col=0; col<35; col++)
{
if (input[data][col] == 1)
System.out.print("*");
else
System.out.print(" ");
if (col==4 || col==9 || col==14 || col==19 || col==24 || col=29 || col=24)
System.out.println();
}
System.out.println();
System.out.println("Target");
for (col=0; col<10; col++)
{
System.out.println(target[data][col]);
}
System.out.println();
}
else
{
System.out.println("Wrong input. Pick a number between 0 - 9");
}
//To stop the program
Scanner scanner = new Scanner(System.in);
System.out.println("Continue? (Y/N) : ");
char ch = scanner.next().charAt(0);
if(ch=='Y' || ch=='y')
{
System.out.println("exiting");
break;
}
}
return 0;
}
static double sigmoid (double a)
{
return 1 / (1 + Math.exp(-(a)));
}
}
编译后我得到了4个错误
error: illegal start of expression
static double sigmoid (double a)
^
error: ';' expected
static double sigmoid (double a)
^
error: ';' expected
static double sigmoid (double a)
^
error: reached end of file while parsing
}
^
4 errors
答案 0 :(得分：2)
在方法声明之前的某处，你有一个缺失的结束}。
答案 1 :(得分：0)
您缺少“sigmoid”方法的访问级别修饰符， 只需在声明的第一个位置键入public，private或protected。