XOR Machine Learning Example (Encog, Java 8 + annotated)

MachineLearningApplication.java

package xyz.jacobclark;

import org.encog.engine.network.activation.ActivationSigmoid;
import org.encog.ml.data.MLData;
import org.encog.ml.data.MLDataPair;
import org.encog.ml.data.basic.BasicMLDataSet;
import org.encog.neural.networks.BasicNetwork;
import org.encog.neural.networks.layers.BasicLayer;
import org.encog.neural.networks.training.propagation.resilient.ResilientPropagation;

public class MachineLearningApplication {
    public static void main(String args[]) {

        /*

            Encog + Java 8

            Machine Learning 3 layer neural network which classifies whether an XOR expression is truthy or falsey in a range of 0d - 1d

            Overview: Neural network with a sigmoid activation function which takes input and ideal outputs through an resilient propagation training algorithm

            Topics Covered:

                Training data setup
                Neural network setup
                Network training
                Running against real values

            Neurals: (simplified for text)

                I1 -(bias)> H1
                                -> O1
                I1 -(bias)> H2

            Activation function:

                Sigmoid

            Training method:

                Rprop (Resilient Propagation)

         */

        // -----
        // Training input

        // Training data set

        // XOR -> Exclusive OR training set
        // Performs ML on the real values of an XOR operation.

        // Example XOR operations and their values
        // 0.0 + 0.0 = False
        // 1.0 + 0.0 = True
        // 0.0 + 1.0 = True
        // 1.0 + 1.0 = False

        double[][] XOR_INPUT = {
                { 0.0, 0.0 },
                { 1.0, 0.0 },
                { 0.0, 1.0 },
                { 1.0, 1.0 }
        };

        // Training input XOR evaluations
        double[][] XOR_IDEAL = {
                {0.0},
                {1.0},
                {1.0},
                {0.0}
        };

        BasicMLDataSet basicMLDataSet = new BasicMLDataSet(XOR_INPUT, XOR_IDEAL);

        // -----
        // Neural Network

        // Create the network
        BasicNetwork network = new BasicNetwork();

        // Create input network layer
        // Creates a layer with 2 neurons and a bias nuron
        BasicLayer input = new BasicLayer(null, true, 2);

        // Sigmoid = Mathematical Function

        // Create hidden network layer
        // Takes the activation sigmoid and creates a layer with 2 neurons and a bias nuron
        BasicLayer hidden = new BasicLayer(new ActivationSigmoid(), true, 2);

        // Create output network layer
        // Takes the activation sigmoid and creates a layer with 1 neuron
        BasicLayer output = new BasicLayer(new ActivationSigmoid(), false, 1);

        // Add network layers
        network.addLayer(input);
        network.addLayer(hidden);
        network.addLayer(output);

        // All networks are added
        network.getStructure().finalizeStructure();

        // Randomly init all weights
        network.reset();

        // -----
        // Training

        // Create training data set (regression)
        ResilientPropagation resilientPropagation = new ResilientPropagation(network, basicMLDataSet);

        // Train the network until the error rate gets very low (1%)
        do {
            resilientPropagation.iteration();
        } while(resilientPropagation.getError() > 0.01);

        // -----
        // Running ideal data against trained network

        // Iterate over each pair of training data created above
        for(MLDataPair pair: basicMLDataSet ) {
            // Compute the real value (hypotehsis function)
            final MLData data = network.compute(pair.getInput());

            // Log out the values
            System.out.println(
                    "Input = "
                    + pair.getInput().getData(0) + ", " + pair.getInput().getData(1)
                    + " | Computed = " + data.getData(0)
                    + " | Ideal = " + pair.getIdeal().getData(0)
            );
        }

        // -----
        // Program output

        /*
            Input = 0.0, 0.0 | Computed = 0.13492055225524496 | Ideal = 0.0
            Input = 1.0, 0.0 | Computed = 0.9480355027515462 | Ideal = 1.0
            Input = 0.0, 1.0 | Computed = 0.9133561421396043 | Ideal = 1.0
            Input = 1.0, 1.0 | Computed = 0.06562084788083862 | Ideal = 0.0
         */
    }
}