README.md

MCMC for Bayesian Non-Parametric Mixture Models

What's this about

This package has been written by Yee Whye Teh for the articles MCMC for Normalized Random Measure Mixture Models and On a class of σ-stable Poisson–Kingman models and an effective marginalized sampler. The folder src/nrmix is related to the first article and src/qmixto the second article. This package has been cleaned and uploaded by Emile Mathieu.

Check also the (partial) Julia port of this package.

Do not hesitate to create pull requests for enhancements or to open issues.

Installation and requirements

Requirements:

• Java JDK [1.8.x] (Can be found on Oracle website)

In a terminal:

git clone https://github.com/emilemathieu/NRMMM.git

Algorithms implemented

• Marginalized Samplers:
  • Neal’s Algorithm 8 generalized
  • The Reuse algorithm
• Conditional Slice Sampler

Example

You can run the Reuse conditional sampler on the galaxy dataset modelled as a mixtures of Gaussian with a normalized generalized Gamma prior:

cd out/artifacts/BNPMix_jar
java -cp BNPMix.jar nrmix.galaxy

Compile project

With IntelliJ:

• File > New > Project from Existing Sources... > Navigate to the NRMMM folder > Open
• File > Project Structure > Project Settings > Artifacts > Click green plus sign > Jar > From modules with dependencies... > Extract to the target Jar > Ok
• Build > Build Artifact...

References

• Favaro, Teh, MCMC for Normalized Random Measure Mixture Models, 2013
• Favaro, Lomeli, Teh, On a class of σ-stable Poisson–Kingman models and an effective marginalized sampler, 2015