Class MetropolisHastings

java.lang.Object

org.episteme.core.mathematics.statistics.montecarlo.MetropolisHastings

public class MetropolisHastings
extends Object

Metropolis-Hastings Markov Chain Monte Carlo (MCMC).

Samples from complex probability distributions. Essential for statistical physics and Bayesian inference.

Since:

1.0

Author:

Silvere Martin-Michiellot, Gemini AI (Google DeepMind)

Constructor Summary

Constructors

Constructor	Description
MetropolisHastings()
MetropolisHastings(long seed)

Method Summary

Modifier and Type	Method	Description
static double	effectiveSampleSize(double[] samples)	Effective sample size (accounts for autocorrelation).
static double	mean(double[] samples)	Compute mean of samples.
double[]	sample1D(Function<Double,Double> logProbability, Function<Double,Double> proposal, double initial, int samples, int burnIn)	Run Metropolis-Hastings sampling.
double[][]	sampleND(Function<double[],Double> logProbability, double[] proposalStd, double[] initial, int samples, int burnIn)	Multi-dimensional Metropolis-Hastings.
static double	variance(double[] samples)	Compute variance of samples.

Methods inherited from class Object

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

Constructor Details

MetropolisHastings

public MetropolisHastings()

MetropolisHastings

public MetropolisHastings(long seed)

Method Details

sample1D

public double[] sample1D(Function<Double,Double> logProbability,
 Function<Double,Double> proposal,
 double initial,
 int samples,
 int burnIn)

Run Metropolis-Hastings sampling.

Parameters:

logProbability - Log probability of target distribution (unnormalized OK)

proposal - Proposal distribution: given current state, generate candidate

initial - Initial state

samples - Number of samples to generate

burnIn - Number of initial samples to discard

Returns:

Array of samples from the target distribution

sampleND

public double[][] sampleND(Function<double[],Double> logProbability,
 double[] proposalStd,
 double[] initial,
 int samples,
 int burnIn)

Multi-dimensional Metropolis-Hastings.

Parameters:

logProbability - Log probability of target distribution

proposalStd - Standard deviation for Gaussian random walk proposal

initial - Initial state vector

samples - Number of samples

burnIn - Burn-in period

Returns:

2D array [sample_index][dimension]

mean

public static double mean(double[] samples)

Compute mean of samples.

variance

public static double variance(double[] samples)

Compute variance of samples.

effectiveSampleSize

public static double effectiveSampleSize(double[] samples)

Effective sample size (accounts for autocorrelation).