org.broadinstitute.hellbender.utils.mcmc

Class SliceSampler

java.lang.Object

org.broadinstitute.hellbender.utils.mcmc.SliceSampler

public final class SliceSampler
extends java.lang.Object

Implements slice sampling of a continuous, univariate, unnormalized probability density function (PDF), which is assumed to be unimodal. See Neal 2003 at https://projecteuclid.org/euclid.aos/1056562461 for details.

Constructor Summary

Constructors

Constructor and Description
SliceSampler(org.apache.commons.math3.random.RandomGenerator rng, java.util.function.Function<java.lang.Double,java.lang.Double> logPDF, double width) Creates a new sampler for an unbounded univariate random variable, given a random number generator, a continuous, univariate, unimodal, unnormalized log probability density function, and a step width.
SliceSampler(org.apache.commons.math3.random.RandomGenerator rng, java.util.function.Function<java.lang.Double,java.lang.Double> logPDF, double xMin, double xMax, double width) Creates a new sampler for a bounded univariate random variable, given a random number generator, a continuous, univariate, unimodal, unnormalized log probability density function, hard limits on the random variable, and a step width.

Method Summary

Modifier and Type	Method and Description
double	sample(double xInitial) Generate a single sample, given an initial value to use in slice construction.
java.util.List<java.lang.Double>	sample(double xInitial, int numSamples) Generate multiple samples from the probability density function, given an initial value to use in slice construction.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SliceSampler

public SliceSampler(org.apache.commons.math3.random.RandomGenerator rng,
                    java.util.function.Function<java.lang.Double,java.lang.Double> logPDF,
                    double xMin,
                    double xMax,
                    double width)

Creates a new sampler for a bounded univariate random variable, given a random number generator, a continuous, univariate, unimodal, unnormalized log probability density function, hard limits on the random variable, and a step width.

Parameters:

rng - random number generator, never null

logPDF - continuous, univariate, unimodal log probability density function (up to additive constant), never null

xMin - minimum allowed value of the random variable

xMax - maximum allowed value of the random variable

width - step width for slice expansion

SliceSampler

public SliceSampler(org.apache.commons.math3.random.RandomGenerator rng,
                    java.util.function.Function<java.lang.Double,java.lang.Double> logPDF,
                    double width)

Creates a new sampler for an unbounded univariate random variable, given a random number generator, a continuous, univariate, unimodal, unnormalized log probability density function, and a step width.

Parameters:

rng - random number generator, never null

logPDF - continuous, univariate, unimodal log probability density function (up to additive constant), never null

width - step width for slice expansion

Method Detail

sample

public double sample(double xInitial)

Generate a single sample, given an initial value to use in slice construction.

Parameters:

xInitial - initial value to use in slice construction; must be in [xMin, xMax]

Returns:

sample drawn from the probability density function

sample

public java.util.List<java.lang.Double> sample(double xInitial,
                                               int numSamples)

Generate multiple samples from the probability density function, given an initial value to use in slice construction.

Parameters:

xInitial - initial value to use in slice construction; if outside [xMin, xMax], forced to be within

numSamples - number of samples to generate

Returns:

samples drawn from the probability density function