Jenetics:柱基因型编解码器的正确使用

Question

目前，我正在研究一个Jenetics (链接到jenetics)实现，以优化粒子加速器束流线。我的健身功能调用加速器检测器设备，定义如下：

private double fitness(final DoubleChromosome chromosomes) {
    // private double fitness(final Genotype<DoubleGene> chromosomes) {
    // Convert genes to a format the device scanner can understand
    // we will inject a 1:n Set<List<Double>>
    final Set<List<Double>> trimValues = new HashSet<>();

    final List<Double> valueList = new ArrayList<>();
    for (final DoubleGene chromosome : chromosomes) {
        valueList.add(Double.valueOf(chromosome.doubleValue()));
    }
    trimValues.add(valueList);

    ....
    more code specific to application
    }

Jenetics的流引擎是在一个特定的方法中初始化的：

public void initAlgorithm(final Object scanParameters) throws Exception {
    if (scanParameters != null) {
        /// See constructor of EvolvingImagesWorker
        _geneticScanParameters = (GeneticScanParameters) scanParameters;
    }

    if (_geneticScanParameters.getTrimParameterSets() != null) {

        final int chromosomeCount = _geneticScanParameters.getTrimParameterSets().size();
        if (chromosomeCount > 0) {
            ISeq<DoubleChromosome> chromosomeSet = ISeq.empty();

            // create an ISeq of genes
            for (final TrimParameterValueSet valueSet : _geneticScanParameters.getTrimParameterSets()) {
                final double minValue = valueSet.getMinValue();
                final double maxValue = valueSet.getMaxValue();
                final double initialValue = (maxValue + minValue) / 2;

                final DoubleGene doubleGene = DoubleGene.of(initialValue, minValue, maxValue);
                final DoubleChromosome doubleChromosome = DoubleChromosome.of(doubleGene.newInstance());
                chromosomeSet = chromosomeSet.append(doubleChromosome.newInstance());
            }
            Codec<DoubleChromosome, DoubleGene> codec = null;
            try {
                final Genotype<DoubleGene> genotype = Genotype.of(chromosomeSet);
                codec = Codec.of(genotype.newInstance(), //
                        gt -> (DoubleChromosome) gt.getChromosome());
            } catch (final IllegalArgumentException ex) {
                MessageLogger.logError(getClass(), Thread.currentThread(), ex);
                throw ex;
            }

            _scannerEngine = Engine.builder(this::fitness, codec) //
                    .executor(Executors.newSingleThreadExecutor()) // without this command, engine will be executed
                                                                   // in
                                                                   // parallel threads
                    .populationSize(_geneticScanParameters.getPopulationSize()) //
                    .optimize(_geneticScanParameters.getOptimizationStrategy()) //
                    .offspringFraction(_geneticScanParameters.getOffspringSize()) //
                    .survivorsSelector(new RouletteWheelSelector<>()) //
                    .offspringSelector(new TournamentSelector<>(_geneticScanParameters.getTournamentSizeLimit())) //
                    .alterers( //
                            new Mutator<>(_geneticScanParameters.getMutator()), //
                            new MeanAlterer<>(_geneticScanParameters.getMeanAlterer()) //
                    ) //
                    .build();
        } else {
            throw new IllegalStateException(ERROR_INITSCANNER_NO_SETTING_DEVICE);
        }
    }
    }

其中：

private Engine<DoubleGene, Double> _scannerEngine = null;

我想要做的是把健身功能称为健身功能，这样我就可以在健身功能中获得基因的值(我发送给加速器的设置)。我已经尝试将健身()定义为：

private double fitness(final Genotype<DoubleChromosome> genotype) {
...
}

但是这个调用会导致遵从性错误。

Answer 1

我看了一下你的代码，我想你想做这样的事情：

class Foo {

    // Your parameter class.
    class TrimParameterSet {
        double min, max;
    }

    static double fitness(final double[] values) {
        // Your fitness function.
        return 0;
    }

    public static void main(final String[] args) {
        final List<TrimParameterSet> valueSets = ...;

        final DoubleRange[] ranges = valueSets.stream()
            .map(p -> DoubleRange.of(p.min, p.max))
            .toArray(DoubleRange[]::new);

        final Codec<double[], DoubleGene> codec = Codecs.ofVector(ranges);
        final Engine<DoubleGene, Double> engine = Engine.builder(Foo::fitness, codec)
            .build();

        // ...
    }
}

适应度函数的double[]数组有一个不同的范围，这与您的TrimParameterSet类中定义的范围不同。如果你想要定义一个直接的适应度函数，你必须定义一个基因型，一个基因作为参数类型。

double fitness(Genotype<DoubleGene> gt) {...}

Answer 2

由于注释函数不允许我完成记录我的最后实现，下面是实际代码：

ISeq<DoubleChromosome> chromosomeSet = ISeq.empty();

// create an ISeq of genes
for (loop criteria) {
    final DoubleGene doubleGene = DoubleGene.of(initialValue, minValue, maxValue);
    final DoubleChromosome doubleChromosome = DoubleChromosome.of(doubleGene.newInstance());
    chromosomeSet = chromosomeSet.append(doubleChromosome.newInstance());
}
_genotype = Genotype.of(chromosomeSet);

_scannerEngine = Engine.builder(this::fitness, _genotype)
                     ... // engine settings
                     .build();

double fitness(Genotype<DoubleGene> gt) {...}