source: java/ecj/cecj/statistics/AverageObjectiveFitnessStatistics.java @ 1326

package cecj.statistics;

import java.io.File;
import java.io.IOException;

import ec.EvolutionState;
import ec.Individual;
import ec.Statistics;
import ec.util.Output;
import ec.util.Parameter;

public class AverageObjectiveFitnessStatistics extends Statistics {

        private static final String P_COUNT_LAST = "count-last";
        private static final String P_FREQUENCY = "frequency";
        private static final String P_FITNESS_CALCULATOR = "fitness-calc";
        private static final String P_FITNESS_FILE = "fitness-file";

        public int fitnessStatisticslog;

        public AverageObjectiveFitnessStatistics() {
                fitnessStatisticslog = 0;
        }

        private int frequency;
        private ObjectiveFitnessCalculator fitnessCalc;

        private int countLast;
        private int lastCounter;
        private float sumLastFitnesses;
        private float lastFitnesses[];

        @Override
        public void setup(EvolutionState state, Parameter base) {
                super.setup(state, base);

                Parameter countLastParam = base.push(P_COUNT_LAST);
                countLast = state.parameters.getIntWithDefault(countLastParam, null, -1);
                if (countLast != -1) {
                        lastFitnesses = new float[countLast];
                        lastCounter = 0;
                        sumLastFitnesses = 0.0f;
                }

                Parameter fitnessCalcParameter = base.push(P_FITNESS_CALCULATOR);
                fitnessCalc = (ObjectiveFitnessCalculator) state.parameters
                        .getInstanceForParameter(fitnessCalcParameter, null, ObjectiveFitnessCalculator.class);
                fitnessCalc.setup(state, fitnessCalcParameter);

                Parameter frequencyParam = base.push(P_FREQUENCY);
                frequency = state.parameters.getIntWithDefault(frequencyParam, null, 1);

                File fitnessStatisticsFile = state.parameters.getFile(base.push(P_FITNESS_FILE), null);
                if (fitnessStatisticsFile != null) {
                        try {
                                fitnessStatisticslog = state.output.addLog(fitnessStatisticsFile, Output.V_VERBOSE,
                                                                                                                        false, true, false);
                        } catch (IOException i) {
                                state.output.fatal("An IOException occurred while trying to create the log "
                                                + fitnessStatisticsFile + ":\n" + i);
                        }
                }
        }

        @Override
        public void postEvaluationStatistics(EvolutionState state) {
                super.postEvaluationStatistics(state);

                if ((state.generation % frequency != 0) && (state.generation != state.numGenerations - 1)) {
                        return;
                }

                for (int subpop = 0; subpop < state.population.subpops.length; ++subpop) {
                        float sumFitness = 0;
                        float maxFitness = Float.NEGATIVE_INFINITY;
                        float maxSubjectiveFitness = Float.NEGATIVE_INFINITY;
                        float objectiveFitnessOfSubjectivelyBest = -1;

                        float[] fitnesses = new float[state.population.subpops[subpop].individuals.length];

                        for (int i = 0; i < state.population.subpops[subpop].individuals.length; ++i) {
                                Individual ind = state.population.subpops[subpop].individuals[i];
                                fitnesses[i] = fitnessCalc.calculateObjectiveFitness(state, ind);

                                sumFitness += fitnesses[i];
                                maxFitness = Math.max(maxFitness, fitnesses[i]);
                                if (ind.fitness.fitness() > maxSubjectiveFitness) {
                                        maxSubjectiveFitness = ind.fitness.fitness();
                                        objectiveFitnessOfSubjectivelyBest = fitnesses[i];
                                }
                        }

                        float sumSquareDiffs = 0;
                        float avgFitness = sumFitness / state.population.subpops[subpop].individuals.length;
                        for (int i = 0; i < state.population.subpops[subpop].individuals.length; ++i) {
                                float diff = fitnesses[i] - avgFitness;
                                sumSquareDiffs += (diff * diff);
                        }

                        double stdev = Math.sqrt(sumSquareDiffs
                                        / state.population.subpops[subpop].individuals.length);

                        if (countLast != -1) {
                                sumLastFitnesses -= lastFitnesses[lastCounter];
                                sumLastFitnesses += objectiveFitnessOfSubjectivelyBest;

                                lastFitnesses[lastCounter] = objectiveFitnessOfSubjectivelyBest;
                                lastCounter++;
                                lastCounter %= countLast;

                                float avgLastFitness = sumLastFitnesses / countLast;
                                state.output.println(state.generation + "\t" + avgFitness + "\t" + stdev + "\t"
                                                + maxFitness + "\t" + objectiveFitnessOfSubjectivelyBest + "\t"
                                                + avgLastFitness, Output.V_VERBOSE + 1, fitnessStatisticslog);
                        } else {
                                state.output.println(state.generation + "\t" + avgFitness + "\t" + stdev + "\t"
                                                + maxFitness + "\t" + objectiveFitnessOfSubjectivelyBest,
                                                                                Output.V_VERBOSE + 1, fitnessStatisticslog);
                        }
                }

                if (state.generation == state.numGenerations - 1) {
                        saveIndividuals(state);
                }
        }

        private void saveIndividuals(EvolutionState state) {
                for (int subpop = 0; subpop < state.population.subpops.length; ++subpop) {
                        state.output.println("\nSubpopulation " + subpop + " individuals : ",
                                                                        Output.V_VERBOSE + 1, fitnessStatisticslog);
                        Individual[] inds = state.population.subpops[subpop].individuals;
                        for (Individual ind : inds) {
                                ind.printIndividual(state, fitnessStatisticslog, Output.V_VERBOSE + 1);
                        }
                }
        }
}