Package com.barrybecker4.optimization.parameter

Examples of com.barrybecker4.optimization.parameter.ParameterArray

51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
     * @return list of moves to a solution.
     */
    @Override
    public List<Piece> solve()  {

        ParameterArray initialGuess = new PieceParameterArray(pieces_);
        solution_ = pieces_;
        long startTime = System.currentTimeMillis();

        Optimizer optimizer = new Optimizer(this);

        optimizer.setListener(this);

        ParameterArray solution =
            optimizer.doOptimization(strategy, initialGuess, MAX_FITS);

        solution_ = ((PieceParameterArray)solution).getPieceList();
        List<Piece> moves;
        if (evaluateFitness(solution) >= MAX_FITS) {
View Full Code Here
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
     * @return the optimized params.
     */
    @Override
    public ParameterArray doOptimization( ParameterArray params, double fitnessRange ) {

        ParameterArray currentParams = params.copy();

        double jumpSize = INITIAL_JUMP_SIZE;

        if (!optimizee_.evaluateByComparison()) {
            // get the initial baseline fitness value.
            currentParams.setFitness(optimizee_.evaluateFitness(currentParams));
        }
        int numIterations = 0;
        log(0, currentParams.getFitness(), 0.0, 0.0, currentParams, "initial test");
        notifyOfChange(currentParams);

        double fitnessEps = fitnessRange * FITNESS_EPS_PERCENT / 100.0;

        // Use cache to avoid repeats. This can be a real issue if  we have a discrete problem space.
        Set<ParameterArray> cache = new HashSet<ParameterArray>();
        cache.add(currentParams);

        Improvement improvement = null;
        boolean improved = true;

        // iterate until there is no significant improvement between iterations,
        // of the jumpSize is too small (below some threshold).
        do {
            //System.out.println( "iter=" + numIterations + " FITNESS = " + currentParams.getFitness() + "  ------------");

            improvement = currentParams.findIncrementalImprovement(optimizee_, jumpSize, improvement, cache);

            numIterations++;
            currentParams = improvement.getParams();
            jumpSize = improvement.getNewJumpSize();
            notifyOfChange(currentParams);
View Full Code Here
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
     * @param previousBest the best solution from the previous iteration
     * @return the new best solution.
     */
    @Override
    protected ParameterArray evaluatePopulation(List<ParameterArray> population, ParameterArray previousBest) {
        ParameterArray bestFitness = previousBest;

        Parallelizer<EvaluationWorker> parallelizer =
                new Parallelizer<EvaluationWorker>();

        List<Runnable> workers = new ArrayList<Runnable>(population.size());

        for (ParameterArray candidate : population) {
            workers.add(new EvaluationWorker(candidate, previousBest));
        }

        // blocks until all Callables are done running.
        parallelizer.invokeAllRunnables(workers);

        for (Runnable worker : workers)  {

            EvaluationWorker eworker = (EvaluationWorker)worker;
            double fitness = eworker.getResult();
            if (fitness > bestFitness.getFitness()) {
                bestFitness = eworker.getCandidate();
            }
        }

        return bestFitness.copy();
    }
View Full Code Here
61
62
63
64
65
66
67
68
69
70
71
    @Override
    public ParameterArray doOptimization( ParameterArray params, double fitnessRange ) {

        List<ParameterArray> samples = params.findGlobalSamples(numSamples_);
        double bestFitness = -Double.MAX_VALUE;
        ParameterArray bestParams = params.copy();

        for (ParameterArray sample : samples) {

            double fitness;
            if (optimizee_.evaluateByComparison())
View Full Code Here
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
            params.setFitness(currentFitness);
        }

        // store the best solution we found at any given temperature iteration and use that as the initial
        // start of the next temperature iteration.
        ParameterArray bestParams = params.copy();
        ParameterArray currentParams;

        do { // temperature iteration (temperature drops each time through)
             currentParams = bestParams;

             do {
                 currentParams = findNeighbor(currentParams, ct, temperature, fitnessRange);

                 if (currentParams.getFitness() > bestParams.getFitness()) {
                     bestParams = currentParams.copy();
                     notifyOfChange(bestParams);
                 }
                 ct++;

             } while (ct < N * currentParams.size() && !isOptimalFitnessReached(currentParams));

             ct = 0;
             // keep Reducing the temperature until it reaches tempMin
             temperature *= TEMP_DROP_FACTOR;

View Full Code Here
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
     */
     private ParameterArray findNeighbor(ParameterArray params, int ct, double temperature, double fitnessRange) {

        //double r = (tempMax_/5.0+temperature) / (8.0*(N/5.0+ct)*tempMax_);
        double r = temperature / ((N + ct) * tempMax_);
        ParameterArray newParams = params.getRandomNeighbor(r);
        double dist = params.distance(newParams);

        double deltaFitness;
        double newFitness;
        if (optimizee_.evaluateByComparison()) {
            deltaFitness = optimizee_.compareFitness(newParams, params);
        }
        else {
            newFitness = optimizee_.evaluateFitness(newParams);
            newParams.setFitness(newFitness);
            deltaFitness = newFitness - params.getFitness();
        }

        double probability = Math.pow(Math.E, tempMax_ * deltaFitness / temperature);
        boolean useWorseSolution = RAND.nextDouble() < probability;
View Full Code Here
39
40
41
42
43
44
45
46
47
48
49
    public void doTest(OptimizationStrategyType optType, Optimizer optimizer,
                       ParameterArray initialGuess, double fitnessRange) {

        pointsList = new PointsList(solutionPosition, EDGE_SIZE);
        ParameterArray solution = null;
        try {
            solution = optimizer.doOptimization(optType, initialGuess, fitnessRange);
        } catch(AbstractMethodError e) {
            // allow continuing if the strategy has simply not been implemented yet.
            e.printStackTrace();
View Full Code Here
40
41
42
43
44
45
46
47
48
49
50
        // 3 sample points along each dimension
        gsStrategy.setSamplingRate(NUM_SAMPLES);

        // first find a good place to start
        // perhaps we should try several of the better results from global sampling.
        ParameterArray sampledParams = gsStrategy.doOptimization(params, fitnessRange);

        OptimizationStrategy strategy = new HillClimbingStrategy(optimizee_);
        strategy.setListener(listener_);
        return strategy.doOptimization(sampledParams, fitnessRange);
    }
View Full Code Here
149
150
151
152
153
154
155
156
157
158
        if (GUIUtil.hasBasicService())
            optimizer = new Optimizer( this );
        else
            optimizer = new Optimizer( this, FileUtil.getHomeDir()+ "performance/liquid/liquid_optimization.txt" );
        Parameter[] params = new Parameter[3];
        ParameterArray paramArray = new NumericParameterArray( params );

        setPaused(false);
        optimizer.doOptimization(OptimizationStrategyType.GENETIC_SEARCH, paramArray, 0.3);
    }
View Full Code Here
96
97
98
99
100
101
102
103
104
105
            optimizer = new Optimizer( this, FileUtil.getHomeDir() +"performance/trebuchet/trebuchet_optimization.txt" );
        Parameter[] params = new Parameter[NUM_PARAMS];
        //params[0] = new Parameter( WAVE_SPEED, 0.0001, 0.02, "wave speed" );
        //params[1] = new Parameter( WAVE_AMPLITUDE, 0.001, 0.2, "wave amplitude" );
        //params[2] = new Parameter( WAVE_PERIOD, 0.5, 9.0, "wave period" );
        ParameterArray paramArray = new NumericParameterArray( params );

        setPaused(false);
        optimizer.doOptimization(  OptimizationStrategyType.GENETIC_SEARCH, paramArray, 0.3);
    }
View Full Code Here
TOP

Related Classes of com.barrybecker4.optimization.parameter.ParameterArray

  • com.barrybecker4.common.app.CommandLineOptions
  • com.barrybecker4.common.concurrency.Worker
  • com.barrybecker4.common.expression.mathexpression.MathExpressionParser
  • com.barrybecker4.common.expression.TreeNode
  • com.barrybecker4.common.geometry.ByteLocation
  • com.barrybecker4.common.geometry.Location
  • com.barrybecker4.common.i18n.LocaleType
  • com.barrybecker4.common.math.cutpoints.AbstractCutPointFinder
  • com.barrybecker4.common.math.function.ArrayFunction
  • com.barrybecker4.common.math.function.ErrorFunction
    • Copyright © 2018 www.massapicom. 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.