edu.stevens.cpe.reservior.readout.CMAES.java Source code

Introduction

Here is the source code for edu.stevens.cpe.reservior.readout.CMAES.java

Source

/*******************************************************************************
 *  Copyright 2013 William Koch
 * 
 *    Licensed under the Apache License, Version 2.0 (the "License");
 *    you may not use this file except in compliance with the License.
 *    You may obtain a copy of the License at
 * 
 *      http://www.apache.org/licenses/LICENSE-2.0
 * 
 *    Unless required by applicable law or agreed to in writing, software
 *    distributed under the License is distributed on an "AS IS" BASIS,
 *    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *    See the License for the specific language governing permissions and
 *    limitations under the License.
 ******************************************************************************/
package edu.stevens.cpe.reservior.readout;

import java.io.ByteArrayOutputStream;
import java.io.PrintStream;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Iterator;

import org.apache.commons.math3.analysis.MultivariateFunction;
import org.apache.commons.math3.linear.Array2DRowRealMatrix;
import org.apache.commons.math3.linear.RealMatrix;
import org.apache.commons.math3.optimization.GoalType;
import org.apache.commons.math3.optimization.PointValuePair;
import org.apache.commons.math3.optimization.direct.CMAESOptimizer;
import org.apache.commons.math3.random.MersenneTwister;
import org.apache.log4j.Logger;
import org.encog.ml.data.MLDataPair;
import org.encog.ml.data.MLDataSet;

import edu.stevens.cpe.reservior.ReserviorException;
import edu.stevens.cpe.reservior.ReservoirNetwork;
import edu.stevens.cpe.reservior.layers.ReadoutFunction;
import edu.stevens.cpe.reservior.layers.SpikingOutput;
import edu.stevens.cpe.reservior.neuron.IFSpikingNeuron;

public class CMAES {
    public static Logger logger = Logger.getLogger(CMAES.class);
    private ReservoirNetwork reservoir;
    private MLDataSet trainingSet;
    private double startValue = 0;
    private double lowerBound = 0;
    private double upperBound = 50;
    private double insigma = .1;
    private int epoch = 0;
    private ReadoutFunction readout;

    double minError = Double.POSITIVE_INFINITY;
    double[] bestWeights;

    public CMAES(ReservoirNetwork reservoir, MLDataSet trainingSet, ReadoutFunction readout) {
        this.reservoir = reservoir;
        this.trainingSet = trainingSet;
        this.readout = readout;
    }

    private final MultivariateFunction fitnessFuntion = new MultivariateFunction() {

        @Override
        public double value(double[] weights) {
            double error = 0;
            updateReadoutWeights(weights);
            for (MLDataPair pair : trainingSet) {
                try {

                    reservoir.input(pair.getInput());
                    double[] output = readout.getOutput().getData();
                    double[] target = pair.getIdeal().getData();
                    //error += compareSpikeTrains(target, output);
                    error += ErrorUtility.computeLinearRegressionError(target, output);
                    //logger.info("error");
                } catch (ReserviorException e) {
                    //Because we are in an implemented class kill this way for now.
                    logger.error(e);
                    System.exit(0);
                }
                //After each time the reservoir sees a training entry reset the reservoir
                reservoir.getReservior().reset();
                readout.reset();

            }
            //error *= error;
            CMAES.this.epoch++;
            //ByteArrayOutputStream baos = new ByteArrayOutputStream();
            //   PrintStream ps = new PrintStream(baos);
            //   ps.printf("%2d \t %5.5f",epoch, error);
            //   logger.info(baos.toString());
            logger.info(epoch + "\t" + error + "\t*" + minError);
            if (error < minError) {
                minError = error;
                bestWeights = weights;
            }
            return error;
        }

    };

    public void updateReadoutWeights(double[] newWeights) {
        //TODO remove neuron hardcoding type
        int index = 0;
        for (int i = 0; i < ((SpikingOutput) readout).getNeurons().length; i++) {
            HashMap<String, Double> weights = ((SpikingOutput) readout).getNeurons()[i].getWeights();
            Iterator<String> it = weights.keySet().iterator();
            while (it.hasNext()) {
                String id = it.next();
                weights.put(id, newWeights[index]);
                index++;
            }
        }
    }

    public double[] train(int numberWeightsToTrain) {
        this.epoch = 0;

        //Number of weights to evolve
        final double[] start = new double[numberWeightsToTrain];
        Arrays.fill(start, startValue);
        final double[] lower = new double[numberWeightsToTrain];
        Arrays.fill(lower, lowerBound);
        final double[] upper = new double[numberWeightsToTrain];
        Arrays.fill(upper, upperBound);

        //Set to 1/3 the initial search volume
        final double[] sigma = new double[numberWeightsToTrain];
        Arrays.fill(sigma, insigma);

        int lambda = 4 + (int) (3. * Math.log(numberWeightsToTrain));
        int maxEvals = CMAESOptimizer.DEFAULT_MAXITERATIONS;
        double stopValue = .01;
        boolean isActive = true; //Chooses the covariance matrix update method.
        int diagonalOnly = 0;
        int checkFeasable = 0;
        final CMAESOptimizer optimizer = new CMAESOptimizer(lambda, sigma, CMAESOptimizer.DEFAULT_MAXITERATIONS,
                stopValue, isActive, diagonalOnly, checkFeasable, new MersenneTwister(), false);

        final PointValuePair result = optimizer.optimize(maxEvals, fitnessFuntion, GoalType.MINIMIZE, start, lower,
                upper);
        logger.info(Arrays.toString(result.getPoint()));
        logger.info("Best weights: " + Arrays.toString(bestWeights));

        updateReadoutWeights(result.getPoint());
        logger.info("done training.");
        return bestWeights;
    }
}