Package eas.simulation.brain.neural

Source Code of eas.simulation.brain.neural.OldNeuroTranslator

/*
* File name:        NeuroTranslator.java (package eas.users.lukas.evolvableNeuroGPM)
* Author(s):        lko
* Java version:     7.0
* Generation date:  11.04.2013 (18:19:44)
*
* (c) This file and the EAS (Easy Agent Simulation) framework containing it
* is protected by Creative Commons by-nc-sa license. Any altered or
* further developed versions of this file have to meet the agreements
* stated by the license conditions.
*
* In a nutshell
* -------------
* You are free:
* - to Share -- to copy, distribute and transmit the work
* - to Remix -- to adapt the work
*
* Under the following conditions:
* - Attribution -- You must attribute the work in the manner specified by the
*   author or licensor (but not in any way that suggests that they endorse
*   you or your use of the work).
* - Noncommercial -- You may not use this work for commercial purposes.
* - Share Alike -- If you alter, transform, or build upon this work, you may
*   distribute the resulting work only under the same or a similar license to
*   this one.
*
* + Detailed license conditions (Germany):
*   http://creativecommons.org/licenses/by-nc-sa/3.0/de/
* + Detailed license conditions (unported):
*   http://creativecommons.org/licenses/by-nc-sa/3.0/deed.en
*
* This header must be placed in the beginning of any version of this file.
*/

package eas.simulation.brain.neural;

import java.util.ArrayList;
import java.util.LinkedList;
import java.util.List;

import org.ejml.data.DenseMatrix64F;
import org.ejml.factory.LinearSolver;
import org.ejml.factory.LinearSolverFactory;

import eas.miscellaneous.StaticMethods;
import eas.simulation.brain.neural.functions.ActivationFunction;
import eas.simulation.brain.neural.functions.ActivationFunctionConstant;
import eas.simulation.brain.neural.functions.ActivationFunctionIdentity;
import eas.simulation.brain.neural.functions.ActivationFunctionLug;
import eas.simulation.brain.neural.functions.InversionNotSupportedException;
import eas.simulation.brain.neural.functions.TransitionFunction;
import eas.simulation.brain.neural.functions.TransitionFunctionWeightedSum;
import eas.startSetup.GlobalVariables;
import eas.startSetup.ParCollection;

/**
* Using the normal constructor to create an object of this class, the
* resulting translator is a universal translator:<BR>
* <BR>
* Neuron 0 - Bias (constant 1).<BR>
* Neuron 1 - Input for sequences (identity).<BR>
* Neuron 2 - Output for translator construction (identity).<BR>
* Link (1 -> 2) - weighted 1.
*
* @author lko
*/
public class OldNeuroTranslator extends GeneralNeuralNetwork {

    private static final long serialVersionUID = 8242169489373650817L;
    public static final int BIAS_NEURON_ID = 0;
    public static final int INPUT_NEURON_ID = 1;
    public static final int OUTPUT_NEURON_ID = 2;

    /**
     * A list of initial values for the neurons that the neurons are set to
     * at the beginning of a translation.
     */
    private ArrayList<Double> neuronStartValuesInputOrOutputOutput = null;
   
    public OldNeuroTranslator() {
        this(GlobalVariables.getPrematureParameters());
    }
   
    public OldNeuroTranslator(ParCollection params) {
        super(params);
        this.setStandardActFct(new ActivationFunctionLug());
        super.addNeuron(BIAS_NEURON_ID, new ActivationFunctionConstant(1), new TransitionFunctionWeightedSum(), Neuron.HIDDEN_NEURON);
        super.addNeuron(INPUT_NEURON_ID, new ActivationFunctionIdentity(), new TransitionFunctionWeightedSum(), Neuron.INPUT_NEURON);
        super.addNeuron(OUTPUT_NEURON_ID, new ActivationFunctionIdentity(), new TransitionFunctionWeightedSum(), Neuron.OUTPUT_NEURON);
        this.setAllowLinksFromOutputs(true);
        this.setAllowLinksToInputs(false);
        this.setAllowRecurrentLinks(true);
        this.addLink(INPUT_NEURON_ID, OUTPUT_NEURON_ID, 1);
    }

    @Override
    public int addNeuron(int neuronType) {
        return super.addNeuron(neuronType);
    }
   
    @Override
    public void removeNeuron(int neuronID) {
        Neuron neu = this.getNeuron(neuronID);
       
        if (neu.isInput() || neu.isOutput()) {
            throw new RuntimeException("Cannot remove input or output neurons from translator (id: " + neuronID + ").");
        }
       
        super.removeNeuron(neuronID);
    }
   
    @Override
    public int addNeuron(int neuronID, ActivationFunction actFctPhi, TransitionFunction trnFctSigma, int neuronType) {
        if (neuronType == Neuron.INPUT_NEURON || neuronType == Neuron.INPUT_OUTPUT_NEURON || neuronType == Neuron.OUTPUT_NEURON) {
            throw new RuntimeException("Cannot insert additional input or output neurons to translator (id: " + neuronID + ").");
        }
        return super.addNeuron(neuronID, actFctPhi, trnFctSigma, neuronType);
    }

    /**
     * All neurons' net output values are set to the given (input or output) values. 
     * The input and ouput values are reset to 0.
     * If the values are null or do not match the neuron count, all is set to 0.
     */
    private void setNeuronsToStartValues() {
        this.resetAllNeuronInputsAndOutputs();
       
        if (this.neuronStartValuesInputOrOutputOutput != null
                && this.neuronStartValuesInputOrOutputOutput.size() == this.getNeuronCount()) {
            this.setAllNeuronOutputsTo(this.neuronStartValuesInputOrOutputOutput);
        }
    }
   
    private void storeCurrentNeuronValuesAsStartValues() {
        this.neuronStartValuesInputOrOutputOutput = new ArrayList<Double>(this.getNeuronCount());
       
        for (int i = 0; i < this.getNeuronCount(); i++) {
            this.neuronStartValuesInputOrOutputOutput.add(this.getNeuron(i).getNetOutput());
        }
    }
   
    private int maxNumOfNeurons = 10;
   
    private int interpretLink(List<Double> sequence, OldNeuroTranslator transNeu, int i) {
        double source;
        double target;
        double weight;
        try { // This is just in case the genome is missing values at the end.
            this.setInput(INPUT_NEURON_ID, sequence.get(i));
            this.propagate();
//            if (i < 10) if (this.getNeuronCount() > 3) StaticMethods.showImage(this.generateNeuroImage(700), "(" + this.getNeuronCount() + " neurons)");
            source = this.getNeuron(OUTPUT_NEURON_ID).getNetOutput() * maxNumOfNeurons;
            this.setInput(INPUT_NEURON_ID, sequence.get(i + 1));
            this.propagate();
//            if (i < 10) if (this.getNeuronCount() > 3) StaticMethods.showImage(this.generateNeuroImage(700), "(" + this.getNeuronCount() + " neurons)");
            target = this.getNeuron(OUTPUT_NEURON_ID).getNetOutput() * maxNumOfNeurons;
            this.setInput(INPUT_NEURON_ID, sequence.get(i + 2));
            this.propagate();
//            if (i < 10) if (this.getNeuronCount() > 3) StaticMethods.showImage(this.generateNeuroImage(700), "(" + this.getNeuronCount() + " neurons)");
            weight = this.getNeuron(OUTPUT_NEURON_ID).getNetOutput();
           
            if (Double.isNaN(source)) {
                source = 0;
            }
            if (Double.isNaN(target)) {
                target = 0;
            }
            if (Double.isNaN(weight)) {
                weight = 0;
            }
           
            NeuralLinkDummyDouble link = new NeuralLinkDummyDouble(source, target, weight, Integer.MAX_VALUE);
            transNeu.applyNeuralLinkSoft(link);
//            if (i < 10) if (this.getNeuronCount() > 3) StaticMethods.showImage(transNeu.generateNeuroImage(700), "(" + transNeu.getNeuronCount() + " neurons)");
        } catch (Exception e) {
            this.getPars().logInfo("Link not inserted (this means probably that the sequence length is not divisible by 3): " + e.getMessage());
            throw e;
        }
        return i;
    }
   
    public OldNeuroTranslator translateGenomeToTranslator(List<Double> sequence) {
        // Preparations of neural network.
        this.setForwardModeAndResetNetOutput(0, true);
       
        OldNeuroTranslator transNeu = new OldNeuroTranslator(this.getPars());
       
        this.setNeuronsToStartValues();
       
        StaticMethods.showImage(this.generateNeuroImage(1000), "");

        for (int i = 0; i < sequence.size(); i += 3) {
            i = interpretLink(sequence, transNeu, i);
        }
       
        // Propagate until nothing changes anymore.
//        double epsilon = 0.001;
//        this.getInputNeuron().setInput(0);
//        HashSet<Double> valuesOccurred = new HashSet<Double>();
//        while (Math.abs(this.getOutputNeuron().getOutput()) > epsilon) {
//            this.propagate();
//            int size = valuesOccurred.size();
//            valuesOccurred.add(this.getOutputNeuron().getOutput());
//            if (size == valuesOccurred.size()) {
//                break;
//            }
//        }
       
        return transNeu;
    }
   
    @SuppressWarnings("unused")
    private void findAndSetAppropriateStartNetOutputs() {
        final int neuroCount = this.getNeuronCount();
        double[][] freeMembers = new double[neuroCount - 2][1];
        double[][] matrix = new double[neuroCount - 2][neuroCount - 2];

        // Set net outputs of output neurons.
        this.getNeuron(OUTPUT_NEURON_ID).setNetOutput(this.getNeuron(OUTPUT_NEURON_ID).getOutput());
       
        // Set net output of bias neuron.
        this.getNeuron(BIAS_NEURON_ID).setNetOutput(this.getNeuron(BIAS_NEURON_ID).getActFct().activationPhi(0));
       
        for (int i = 2; i < neuroCount; i++) {
            freeMembers[i - 2][0] = this.getNeuron(i).getNetOutput();
            try {
                freeMembers[i - 2][0] = this.getNeuron(i).getActFct().inverseActivationPhi(freeMembers[i - 2][0]);
            } catch (InversionNotSupportedException e) {
                throw new RuntimeException("All non-bias neurons should be invertible.");
            }
        }
       
        for (int i = 0; i < neuroCount - 2; i++) {
            freeMembers[i][0] -= this.getNeuron(BIAS_NEURON_ID).getNetOutput() * this.getWeight(BIAS_NEURON_ID, i + 2);
            freeMembers[i][0] -= this.getNeuron(OUTPUT_NEURON_ID).getNetOutput() * this.getWeight(OUTPUT_NEURON_ID, i + 2);
           
            matrix[i][0] = this.getWeight(1, i + 2);
            for (int j = 3; j < neuroCount; j++) {
                matrix[i][j - 2] = this.getWeight(j, i + 2);
            }
        }
       
        DenseMatrix64F mx = this.solveLinearEquationSystem(matrix, freeMembers);
       
        this.getNeuron(INPUT_NEURON_ID).setNetOutput(this.getNeuron(INPUT_NEURON_ID).getActFct().activationPhi(mx.get(0, 0)));
        for (int i = 1; i < neuroCount - 2; i++) {
            this.getNeuron(i + 2).setNetOutput(this.getNeuron(i + 2).getActFct().activationPhi(mx.get(i, 0)));
        }
       
//        StaticMethods.showImage(this.generateNeuroImage(600), "");
//        try {Thread.sleep(1000000000);} catch (InterruptedException e) {}
    }

    public LinkedList<Double> translateReverseTranslatorToGenome() {
        LinkedList<Double> genomeInput = new LinkedList<Double>();
        LinkedList<NeuralLinkDummyDouble> givenOutput = this.getConstructionSoftLinkHistory();
        ArrayList<Double> givenOutputSequence = new ArrayList<Double>(givenOutput.size());
       
        // Generate sequence of doubles.
        for (NeuralLinkDummyDouble n : givenOutput) {
            givenOutputSequence.add(0, (double) n.getSourceID());
            givenOutputSequence.add(0, (double) n.getTargetID());
            givenOutputSequence.add(0, n.getWeight());
        }

        // Preparations of neural network.
        this.setForwardModeAndResetNetOutput(0, false);
        this.setOutput(OldNeuroTranslator.OUTPUT_NEURON_ID, givenOutputSequence.get(0));
//        this.findAndSetAppropriateStartNetOutputs();
       
        /**
         * Traverse sequence and retranslate to sequence.
         */
        for (int i = 0; i < givenOutputSequence.size(); i++) {
            double value = givenOutputSequence.get(i);
            this.setOutput(OldNeuroTranslator.OUTPUT_NEURON_ID, value);
            this.propagateBackwards();
            genomeInput.add(this.getNeuron(INPUT_NEURON_ID).getInput());
        }
       
        this.storeCurrentNeuronValuesAsStartValues();
        return genomeInput;
    }
   
    public Neuron getOutputNeuron() {
        return this.getNeuron(2);
    }
   
    public Neuron getInputNeuron() {
        return this.getNeuron(1);
    }
   
    private double[][] freeMembers; // Only one column!
    private double[][] matrix;
   
    /**
     * Propagates the net one step backwards by solving a system of linear
     * equations. For now, only the version with neuron0 = bias, neuron1
     * =input (no other incoming signals) and neuron2=output is implemented.
     *
     * @return  The value of the input neuron.
     */
    public double propagateBackwards() {
        final int neuroCount = this.getNeuronCount();
       
        this.generateLinearEquationSystem();
        DenseMatrix64F mx = this.solveLinearEquationSystem(this.matrix, this.freeMembers);

//        StaticMethods.showImage(this.generateNeuroImage(700), "(" + this.getNeuronCount() + " neurons)");
//        try {Thread.sleep(1000000000);} catch (InterruptedException e) {}
       
        // Set neurons at accurate values.
       
        this.getNeuron(INPUT_NEURON_ID).setNetOutput(mx.get(2, 0));
        try {
            this.getNeuron(INPUT_NEURON_ID).setInput(this.getNeuron(INPUT_NEURON_ID).getActFct().inverseActivationPhi(mx.get(2, 0)));
        } catch (InversionNotSupportedException e) {
            throw new RuntimeException("All non-bias neurons should be invertible.");
        }
       
        for (int j = 0; j < neuroCount; j++) {
            this.getNeuron(j).setNetOutput(mx.get(j, 0));
        }
       
        return this.getNeuron(INPUT_NEURON_ID).getInput();
    }

    /**
     * Generates the system of linear equations required to set the current
     * state of the neural network one step backwards in time.<BR>
     * <BR>
     * The system is stored in class variables that should never be touched
     * from outside this method.
     */
    private void generateLinearEquationSystem() {
        final int neuroCount = this.getNeuronCount();
        this.freeMembers = new double[neuroCount - 2][1];
        this.matrix = new double[neuroCount - 2][neuroCount];

        // Set net outputs of output neurons.
        this.getNeuron(OUTPUT_NEURON_ID).setNetOutput(this.getNeuron(OUTPUT_NEURON_ID).getOutput());
       
        // Set net output of bias neuron.
        this.getNeuron(BIAS_NEURON_ID).setNetOutput(this.getNeuron(BIAS_NEURON_ID).getActFct().activationPhi(0));
       
        for (int i = 2; i < neuroCount; i++) {
            freeMembers[i - 2][0] = this.getNeuron(i).getNetOutput();
            try {
                freeMembers[i - 2][0] = this.getNeuron(i).getActFct().inverseActivationPhi(freeMembers[i - 2][0]);
            } catch (InversionNotSupportedException e) {
                throw new RuntimeException("All non-bias neurons should be invertible.");
            }
        }
       
        for (int i = 0; i < neuroCount - 2; i++) {
            for (int j = 0; j < i + 2; j++) {
                freeMembers[i][0] -= this.getNeuron(j).getNetOutput() * this.getWeight(j, i + 2);
            }
           
            for (int j = i + 2; j < neuroCount; j++) {
                matrix[i][j] = this.getWeight(j, i + 2);
            }
        }
    }

    private DenseMatrix64F solveLinearEquationSystem(double[][] matrix, double[][] freeMembers) {
        DenseMatrix64F mA = new DenseMatrix64F(matrix);
        DenseMatrix64F mb = new DenseMatrix64F(freeMembers);
        DenseMatrix64F mx = new DenseMatrix64F(mA.numCols, 1);
       
        LinearSolver<DenseMatrix64F> solver = LinearSolverFactory.leastSquaresQrPivot(true, false);
        solver.setA(mA);
        solver.solve(mb, mx);
       
//        System.out.println(mA);
//        System.out.println(mb);
//        System.out.println(mx);

        return mx;
    }
}
TOP

Related Classes of eas.simulation.brain.neural.OldNeuroTranslator

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.