cern.jet.random.engine.MersenneTwister Java Examples

/**
 * Create n  permutations
 *
 */
private  void init() {



    if (a==null) {
    //    System.out.println("this.seed = "+this.seed);
        MersenneTwister random = new MersenneTwister((int) this.seed);
  a = new long[n];
  b = new long[n];
  for (int j=0; j<n; j++) {
      a[j] = random.nextLong();
      b[j] = random.nextLong();
 //     System.out.println("a["+j+"]="+a[j]);
  //    System.out.println("b["+j+"]="+b[j]);
      while (a[j]<0)  a[j] = random.nextLong();
      while (b[j]<0)  b[j] = random.nextLong();
  }  
    }
}
 

public IndependentNormalDistributionSampler(Variable variable, NormalDistributionModel model, double weight, boolean updateAllIndependently) {
	
	this.variable = variable;
	this.model = model;
	this.updateAllIndependently = updateAllIndependently;
	setWeight(weight);
	
	if (TRY_COLT) {
           randomEngine = new MersenneTwister(MathUtils.nextInt());
           //create standard normal distribution, internal states will be bypassed anyway
           coltNormal = new Normal(0.0, 1.0, randomEngine);
       } else {
       	//no random draw with specified mean and stdev implemented in the normal distribution in BEAST (as far as I know)
       	throw new RuntimeException("Normal distribution in BEAST still needs a random sampler.");
       }
	
}
 

public GibbsIndependentNormalDistributionOperator(Variable variable, NormalDistributionModel model, double weight, boolean updateAllIndependently) {
	
	this.variable = variable;
	this.model = model;
	this.updateAllIndependently = updateAllIndependently;
	setWeight(weight);
	
	if (TRY_COLT) {
           randomEngine = new MersenneTwister(MathUtils.nextInt());
           //create standard normal distribution, internal states will be bypassed anyway
           //takes mean and standard deviation
           coltNormal = new Normal(0.0, 1.0, randomEngine);
       } else {
       	//no random draw with specified mean and stdev implemented in the normal distribution in BEAST (as far as I know)
       	throw new RuntimeException("Normal distribution in BEAST still needs a random sampler.");
       }
	
}
 

public GibbsIndependentJointNormalGammaOperator(Variable mean, Variable precision, NormalDistributionModel model, GammaDistribution gamma, double weight, boolean updateAllIndependently) {
	
	this.mean = mean;
       this.precision = precision;
	this.model = model;
       this.gamma = gamma;
	this.updateAllIndependently = updateAllIndependently;
	setWeight(weight);
	
	if (TRY_COLT) {
           randomEngine = new MersenneTwister(MathUtils.nextInt());
           //create standard normal distribution, internal states will be bypassed anyway
           //takes mean and standard deviation
           coltNormal = new Normal(0.0, 1.0, randomEngine);
           //coltGamma = new Gamma(gamma.getShape(), 1.0/gamma.getScale(), randomEngine);
       } else {
       	//no random draw with specified mean and stdev implemented in the normal distribution in BEAST (as far as I know)
       	throw new RuntimeException("Normal distribution in BEAST still needs a random sampler.");
       }
	
}
 

/** Creates a new Bloom filter with given number of hash functions and expected number of elements.
 * 
 * @param n the expected number of elements.
 * @param d the number of hash functions; if the filter add not more than <code>n</code> elements,
 * false positives will happen with probability 2<sup>-<var>d</var></sup>.
 */
public IntBloomFilter( final int n, final int d ) {
	this.d = d;
	bits = LongArrayBitVector.getInstance().length( (long)Math.ceil( ( n * d / NATURAL_LOG_OF_2 ) ) );
	m = bits.length() * Long.SIZE;

	if ( DEBUG ) System.err.println( "Number of bits: " + m );
	
	// The purpose of Random().nextInt() is to generate a different seed at each invocation.
	final MersenneTwister mersenneTwister = new MersenneTwister( new Random().nextInt() );
	a = new int[ d ];
	b = new int[ d ];
	for( int i = 0; i < d; i++ ) {
		a[ i ] = mersenneTwister.nextInt();
		b[ i ] = mersenneTwister.nextInt();
	}
}
 

/** Creates a new Bloom filter with given number of hash functions and expected number of elements.
 * 
 * @param n the expected number of elements.
 * @param d the number of hash functions; under obvious uniformity and indipendence assumptions,
 * if the filter has not more than <code>n</code> elements,
 * false positives will happen with probability 2<sup>-<var>d</var></sup>.
 */
public BloomFilter( final int n, final int d ) {
	this.d = d;
	final long wantedNumberOfBits = (long)Math.ceil( n * ( d / NATURAL_LOG_OF_2 ) );
	if ( wantedNumberOfBits > MAX_BITS ) throw new IllegalArgumentException( "The wanted number of bits (" + wantedNumberOfBits + ") is larger than " + MAX_BITS );
	bits = new long[ (int)( ( wantedNumberOfBits + Long.SIZE - 1 ) / Long.SIZE ) ];
	m = bits.length * (long)Long.SIZE;

	if ( DEBUG ) System.err.println( "Number of bits: " + m );
	
	// The purpose of Random().nextInt() is to generate a different seed at each invocation.
	final MersenneTwister mersenneTwister = new MersenneTwister( new Random().nextInt() );
	weight = new int[ d ][];
	init = new int[ d ];
	for( int i = 0; i < d; i++ ) {
		weight[ i ] = new int[ NUMBER_OF_WEIGHTS ];
		init[ i ] = mersenneTwister.nextInt();
		for( int j = 0; j < NUMBER_OF_WEIGHTS; j++ )
			 weight[ i ][ j ] = mersenneTwister.nextInt();
	}
}
 

@Test
@Ignore
public void testCorrectArmChosen() {
  int correct = 0;
  for (int i = 0; i< 10000; i++) {
    RandomEngine engine = new MersenneTwister(i);
    BatchedThompsonSampling batchedBandit = new BatchedThompsonSampling();
    batchedBandit.setRandomEngine(engine);
    BatchedBanditTester tester = new BatchedBanditTester(batchedBandit, engine);
    if (i % 100 == 0) {
      System.out.println("Batches complete " + i);
    }
    correct += tester.getWinningArm();
  }
  System.out.println(correct);
  assertTrue(correct > 9500);
}
 

@Test
public void testCorrectArmChosen() {
  int correct = 0;
  for (int i = 0; i<= 10000; i++) {
    RandomEngine engine = new MersenneTwister(i);
    BanditPerformance performance = new BanditPerformance(2);
    BatchedABTest batchedBandit = new BatchedABTest();
    batchedBandit.setRandomEngine(engine);
    BatchedBanditTester tester = new BatchedBanditTester(batchedBandit, engine);
    if (i % 100 == 0) {
      System.out.println("Batches complete " + i);
    }
    correct += tester.getWinningArm();
  }
  assertTrue(correct > 9500);
}
 

/**
   * Creates a new exponential number generator.
   * 
   * @param mean the mean for the distribution.
   */
  public PoissonDistr(double mean) {
engine = new MersenneTwister(new Date());
poisson = new Poisson(mean, engine);

//always sleep for some milliseconds in order not to have same seed for iterative PoissonDistr contruction
try {
	TimeUnit.MILLISECONDS.sleep(10);
} catch (InterruptedException e) {
   	SimLogger.printLine("impossible is occured! Poisson random number cannot be created!");
	e.printStackTrace();
   	System.exit(0);
}
  }
 

/**
 * Constructor
 */
public GammaDistribution(double shape, double scale) {
    this.shape = shape;
    this.scale = scale;
    this.samples = 0;

    if (TRY_COLT) {
        randomEngine = new MersenneTwister(MathUtils.nextInt());
        System.out.println("Colt Gamma(" + shape + "," + scale + ")");
        coltGamma = new Gamma(shape, 1.0/scale, randomEngine);
    }
}