artifisial neural network, menggunakan sebuah layer dan bias.

ANN.java

package com.masasdani.ann;

import java.util.Random;

public class ANN {

  private double alpha; //nilai yg ditentukan di awal, untuk menentukan delta w2
	private int hiddenLayer; // digunakan untuk menentukan banyak node hidden di layer 1
	private double w1[][]; // bobot dari input ke hidden layer
	private double wbias[]; // bias dari input ke hiden layer
	private double wbias2; // bias dari hiddenlayer ke output Y
	private double z[];// nilai hidden layer
	private double w2[]; // bobot dari hidden layer ke output Y
	private double y; //output Y
	private boolean generateRandomWeight = false;
	
	public ANN(double alpha, int hiddenLayerCount) {
		this.alpha= alpha;
		this.hiddenLayer = hiddenLayerCount;
	}
	
	public ANN(double w1[][] , double w2[], double wbias[], double wbias2, int hiddenLayerCount, double alpha){
		this.w1 = w1;
		this.w2 = w2;
		this.alpha= alpha;
		this.hiddenLayer = hiddenLayerCount;
		this.wbias = wbias;
		this.wbias2 = wbias2;
		generateRandomWeight = true;
	}
	
	public void training(double data[], int epoch, double errormin, double target){
		//inisialisasi
		w1 = new double[data.length][hiddenLayer];
		wbias = new double[hiddenLayer];
		z = new double[hiddenLayer];
		w2 = new double[hiddenLayer];
		double sigmaWX[] = new double[hiddenLayer];
		double sigmaWsignoidZ =0d;
		y=0d;
		//generate random bobot
		if(generateRandomWeight) generateRandomWeight(data);
		//lakukan perulangan sebanyak epoch
		for(int x=0;x<epoch;x++){
			System.out.println("perulangan ke = "+x);
			for(int i=0;i<hiddenLayer;i++){
				for(int j=0;j<data.length;j++){
					sigmaWX[i] += w1[j][i];
				}
				z[i] = wbias[i]+sigmaWX[i];
			}
			for(int i=0;i<hiddenLayer;i++){
				sigmaWsignoidZ += sigmoid(z[i]);
			}
			y = wbias2 + sigmaWsignoidZ;
			double sigmoidY = sigmoid(y);
			double errorY = (target-sigmoidY)*sigmoidY*(1-sigmoidY);
			System.out.println(errorY+"");
			/////// fungsi berhenti ketika error y < error minimal
			if(errorY < errormin) 
				return;
			/////// jika tidak, lanjut ke dalam
			double deltaW1[][] = new double[data.length][hiddenLayer];
			double deltaW2[] = new double[hiddenLayer];
			double errornet[] = new double[hiddenLayer];
			double error[] = new double[hiddenLayer];
			for(int i=0;i<hiddenLayer;i++){
				deltaW2[i]=alpha*errorY*sigmoid(z[i]);
				errornet[i] = sigmoidY*deltaW2[i];
				error[i] = errornet[i]*sigmoid(z[i])*(1-sigmoid(z[i]));
			}
			for(int i=0;i<data.length;i++){
				for(int j=0;j<hiddenLayer;j++){
					deltaW1[i][j] = alpha*error[j]*data[i];
				}
			}
			for(int i=0;i<data.length;i++){
				for(int j=0;j<hiddenLayer;j++){
					w1[i][j] = w1[i][j]+deltaW1[i][j];
				}
			}
			for(int i=0;i<hiddenLayer;i++){
				w2[i] = w2[i] + deltaW2[i];
			}
		}
	}
	
	public double[][] getW1(){
		return w1;
	}
	
	public double[] getW2(){
		return w2;
	}
	
	// fungsi aktifasi (sigmoid)
	private double sigmoid(double y){
		return 1 /(1+(Math.pow(Math.E, -y)));
	}
	
	private void generateRandomWeight(double data[]){
		Random random = new Random();
		//inisialisasi bias 1 0.5
		
		for(int i=0;i<wbias.length;i++) wbias[i]=0.5;
		//inisialisasi bias 2 0.1
		wbias2 = 0.1;
		//inisialisasi bobot w1 secara random
		for(int i=0;i<data.length;i++){
			System.out.print("bias "+i+" ");
			for(int j=0;j<hiddenLayer;j++){
				w1[i][j] = random.nextDouble();
				System.out.print(w1[i][j]+" ");
			}
			System.out.println("");
		}
		//inisialisasi bobot w2 secara random
		for(int i=0;i<hiddenLayer;i++){
			w2[i] = random.nextDouble();
			System.out.println("bias "+i+" "+w2[i]);
		}
	}
}

test training menentukan bobot

public static void main(String args[]){
double[] data = {0.8, 0.7};
ANN ann = new ANN(0.3, 2);
ann.training(data, 1000, 0.0005, 3.25);
double w1[][] = ann.getW1();
double w2[] = ann.getW2();

 for(int i=0;i<w1.length;i++){
     for(int j=0;j< w1[i].length;j++){
         System.out.print(w1[i][j]+" ");
     }
     System.out.println();
 }
 for(int i=0; i< w2.length;i++){
     System.out.println(w2[i]);
 }

}