Examples of org.uncommons.maths.random.MersenneTwisterRNG

org.uncommons.maths.random.MersenneTwisterRNG
ath.sci.hiroshima-u.ac.jp/~m-mat/MT/emt.html" target="_top">Mersenne Twister algorithm developed by Makoto Matsumoto and Takuji Nishimura.

This is a very fast random number generator with good statistical properties (it passes the full DIEHARD suite). This is the best RNG for most experiments. If a non-linear generator is required, use the slower {@link AESCounterRNG} RNG.
This PRNG is deterministic, which can be advantageous for testing purposes since the output is repeatable. If multiple instances of this class are created with the same seed they will all have identical output.

This code is translated from the original C version and assumes that we will always seed from an array of bytes. I don't pretend to know the meanings of the magic numbers or how it works, it just does.

NOTE: Because instances of this class require 128-bit seeds, it is not possible to seed this RNG using the {@link #setSeed(long)} method inheritedfrom {@link Random}. Calls to this method will have no effect. Instead the seed must be set by a constructor.
@author Makoto Matsumoto and Takuji Nishimura (original C version) @author Daniel Dyer (Java port)

   * Make sure that evaluate() is called as many different populations are
   * passed to getFitness().
   */
  public void testEvaluateDifferentPopulations() {
    STFitnessEvaluatorMock<Integer> mock = new STFitnessEvaluatorMock<Integer>();
    Random rng = new MersenneTwisterRNG();


    // generate a population A
    int size = 100;
    List<Integer> population = randomInts(size, rng);


    List<Double> evaluations = randomFloats(size, rng);
    mock.shouldReturn(evaluations);


    // call with population A
    mock.getFitness(population.get(rng.nextInt(size)), population);


    // generate a new population B
    population = randomInts(size, rng);


    // call with population B
    mock.getFitness(population.get(rng.nextInt(size)), population);


    // getFitness() should be called twice
    assertEquals(2, mock.getNbCalls());
  }

View Full Code Here

  public static void useTestSeed() {
    testSeed = true;
  }


  public static Random getRandom() {
    return testSeed ? new MersenneTwisterRNG(STANDARD_SEED) : new MersenneTwisterRNG();
  }

View Full Code Here

    seedBytes[4] = (byte) (seed >>> 24);
    seedBytes[5] = (byte) (seed >>> 16);
    seedBytes[6] = (byte) (seed >>>  8);
    seedBytes[7] = (byte) seed;
    System.arraycopy(seedBytes, 0, seedBytes, 8, 8);
    return new MersenneTwisterRNG(seedBytes);
  }

View Full Code Here


  public void testTrainingTestingSets() throws IOException {
    int n = 20;


    for (int nloop = 0; nloop < n; nloop++) {
      MersenneTwisterRNG rng = (MersenneTwisterRNG) RandomUtils.getRandom();
      byte[] seed = rng.getSeed();
      double threshold = rng.nextDouble();


      JobConf conf = new JobConf();
      Set<Long> dataset = new HashSet<Long>();
      LongWritable key = new LongWritable();
      Text value = new Text();

View Full Code Here


  public void testStoreJobParameters() {
    int n = 20;


    for (int nloop = 0; nloop < n; nloop++) {
      MersenneTwisterRNG rng = (MersenneTwisterRNG) RandomUtils.getRandom();


      byte[] seed = rng.getSeed();
      double threshold = rng.nextDouble();
      boolean training = rng.nextBoolean();


      DatasetSplit split = new DatasetSplit(seed, threshold);
      split.setTraining(training);


      JobConf conf = new JobConf();

View Full Code Here

    this.threshold = threshold;
    this.training = true;
  }


  public DatasetSplit(double threshold) {
    this(new MersenneTwisterRNG().getSeed(), threshold);
  }

View Full Code Here


      this.reader = reader;


      DatasetSplit split = new DatasetSplit(conf);


      rng = new MersenneTwisterRNG(split.getSeed());
      threshold = split.getThreshold();
      training = split.isTraining();
    }

View Full Code Here

        TreeEvaluator evaluator = new TreeEvaluator(data);
        EvolutionEngine<Node> engine = new GenerationalEvolutionEngine<Node>(factory,
                                                                             new EvolutionPipeline<Node>(operators),
                                                                             evaluator,
                                                                             new RouletteWheelSelection(),
                                                                             new MersenneTwisterRNG());
        engine.addEvolutionObserver(new EvolutionLogger<Node>());
        return engine.evolve(1000, 5, new TargetFitness(0d, evaluator.isNatural()));
    }

View Full Code Here

                                                                                              populationSize,
                                                                                              1);
                EvolutionEngine<Biomorph> engine = new GenerationalEvolutionEngine<Biomorph>(new BiomorphFactory(),
                                                                                             mutation,
                                                                                             selection,
                                                                                             new MersenneTwisterRNG());
                engine.addEvolutionObserver(new GenerationTracker());
                return engine.evolve(populationSize,
                                     0,
                                     new GenerationCount(generationCount));
            }

View Full Code Here

        EvolutionaryOperator<BitString> pipeline = new EvolutionPipeline<BitString>(operators);
        GenerationalEvolutionEngine<BitString> engine = new GenerationalEvolutionEngine<BitString>(new BitStringFactory(length),
                                                                                                   pipeline,
                                                                                                   new BitStringEvaluator(),
                                                                                                   new RouletteWheelSelection(),
                                                                                                   new MersenneTwisterRNG());
        engine.setSingleThreaded(true); // Performs better for very trivial fitness evaluations.
        engine.addEvolutionObserver(new EvolutionLogger<BitString>());
        return engine.evolve(100, // 100 individuals in each generation.
                             0, // Don't use elitism.
                             new TargetFitness(length, true)); // Continue until a perfect match is found.

View Full Code Here

0 1 2 3

TOP

Related Classes of org.uncommons.maths.random.MersenneTwisterRNG

com.github.neuralnetworks.training.random.MersenneTwisterRandomInitializer

org.apache.mahout.clustering.dirichlet.UncommonDistributions

org.apache.mahout.common.RandomUtils

org.apache.mahout.common.RandomWrapper

org.apache.mahout.ga.watchmaker.cd.CDCrossoverTest

org.apache.mahout.ga.watchmaker.cd.CDFitnessTest

org.apache.mahout.ga.watchmaker.cd.CDGA

org.apache.mahout.ga.watchmaker.cd.CDMutationTest

org.apache.mahout.ga.watchmaker.cd.CDRuleTest

org.apache.mahout.ga.watchmaker.cd.hadoop.CDMahoutEvaluatorTest

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.