net.sourceforge.cilib.stoppingcondition.NNPerformanceChangeStoppingCondition Maven / Gradle / Ivy

Go to download
/**           __  __
*    _____ _/ /_/ /_    Computational Intelligence Library (CIlib)
*   / ___/ / / / __ \   (c) CIRG @ UP
*  / /__/ / / / /_/ /   http://cilib.net
*  \___/_/_/_/_.___/
*/
package net.sourceforge.cilib.stoppingcondition;

import java.util.ArrayList;
import net.sourceforge.cilib.algorithm.Algorithm;
import net.sourceforge.cilib.io.pattern.StandardPattern;
import net.sourceforge.cilib.io.StandardPatternDataTable;
import net.sourceforge.cilib.nn.NeuralNetwork;
import net.sourceforge.cilib.problem.nn.NNTrainingProblem;
import net.sourceforge.cilib.stoppingcondition.nnperformancecomparators.NNPerformanceComparator;
import net.sourceforge.cilib.stoppingcondition.nnperformancecomparators.OneTailedWilcoxonNNPerformanceComparator;
import net.sourceforge.cilib.type.types.container.Vector;

/**
 * Algorithm is complete if no significant improvement in NN performance is observed.
 * Statistical tests such as the (Wilcoxon signed-rank test) can be used to determine
 * the significance of an improvement.
 * 
 * This class keeps memory of the performance of the NN from previous stopping condition test.
 */
public class NNPerformanceChangeStoppingCondition implements StoppingCondition {

    protected ArrayList previousResults;
    private NNPerformanceComparator comparator;

    public NNPerformanceChangeStoppingCondition() {
        previousResults = null;
        comparator = new OneTailedWilcoxonNNPerformanceComparator();
    }

    public NNPerformanceChangeStoppingCondition(NNPerformanceChangeStoppingCondition rhs) {
        previousResults = rhs.previousResults;
        comparator = rhs.comparator;
    }
    
    /**
     * {@inheritDoc}
     */
    @Override
    public NNPerformanceChangeStoppingCondition getClone() {
        return new NNPerformanceChangeStoppingCondition(this);
    }

    /**
     * {@inheritDoc }
     */
    @Override
    public double getPercentageCompleted(Algorithm algorithm) {
        return Math.min(1.0, comparator.getLastPValue()/(comparator.getTargetPValue()));
    }

    /**
     * Applies stopping condition test.
     * @param algorithm The algorithm to be stopped.
     * @result true if algorithm is complete.
     */
    @Override
    public boolean apply(Algorithm algorithm) {
        NNTrainingProblem problem = (NNTrainingProblem) algorithm.getOptimisationProblem();
        StandardPatternDataTable validationSet = problem.getValidationSet();
        NeuralNetwork currentNetwork = problem.getNeuralNetwork();
        currentNetwork.setWeights((Vector) algorithm.getBestSolution().getPosition());

        //calculate output set for the new architecture
        ArrayList currentResults = new ArrayList();
        for (StandardPattern curPattern : validationSet) {
            Vector output = currentNetwork.evaluatePattern(curPattern);
            currentResults.add(output);
        }

        boolean isSame = false;
        if (previousResults != null)
            isSame = comparator.isSame(validationSet, previousResults, currentResults);

        previousResults = currentResults;

        return isSame;
    }

    /**
     * Sets the NN comparator to be used.
     * @param newComparator The new comparator.
     */
    public void setComparator(NNPerformanceComparator newComparator) {
        comparator = newComparator;
    }

    /**
     * Gets the results from the new NN from the previous test.
     * Should only be used for testing purposes.
     * @return Previous results.
     */
    public ArrayList getPreviousResults() {
        return previousResults;
    }
}