/** A simple Perceptron example. No extra comments since the code is self explanatory. */ import java.util.Arrays; import java.util.List; import java.util.ArrayList; import java.util.Random; public class Program { public static void main(String[] args) { Perceptron.demo(); } } class Perceptron{ private List inputs; public List weights; public double bias = 1.; public double activation; private double target; private double delta; public double error; private double learningRate = 0.1; private double momentum = .1; public Perceptron(){ weights = generateWeights(2); } public static void demo(){ Perceptron perceptron = new Perceptron(); List> inputs = new ArrayList<>(); List targets = new ArrayList<>(); targets.add(1.); targets.add(0.); targets.add(0.); targets.add(1.); Double[][] arrays = {{1., 1.}, {1., 0.}, {0., 1.}, {0., 0.}}; int epochs = 20; for(Double[] arr: arrays){ inputs.add(Arrays.asList(arr)); } System.out.println("The perceptron seems to coverge after about 14 updates...\n"); for(int i = 0; i < epochs; i ++){ for(int j = 0; j < inputs.size(); j++){ perceptron.forwardPropagate(inputs.get(j), targets.get(j)); perceptron.backpropagate(); } if(i % 2 != 0) continue; System.out.println(String.format("Epoch = %d Error = %f", i, perceptron.error)); } } public double forwardPropagate(List inputs, double target){ this.inputs = inputs; this.target = target; activation = sigmoid(dotProduct(weights, inputs) + bias); error = 0.5 * Math.pow(target - activation, 2); double errorDerivative = 2 * (activation - target); delta = (sigInverse(activation) * errorDerivative); return activation; } public void backpropagate(){ for(int i = 0; i < weights.size(); i++){ weights.set(i, weights.get(i) + (delta * learningRate) + momentum); bias += (delta * learningRate) + momentum; } } private List generateWeights(int n){ List weights = new ArrayList<>(); Random random = new Random(); for(int i=0; i list1, List list2){ double res = 0.; for(int i = 0; i < list1.size(); i++){ res += list1.get(i) * list2.get(i); } return res; } }