Home

Awesome

NPBayesHMM : Nonparametric Bayesian HMM toolbox, for Matlab

Website: http://michaelchughes.github.com/NPBayesHMM/

Author: Mike Hughes (www.michaelchughes.com, mike AT michaelchughes.com)

This toolbox provides code for running Markov chain Monte Carlo (MCMC) posterior inference for the Beta Process (BP) hidden markov model (HMM).

This software is released under the BSD 3-clause license. Please see the LICENSE file for details.

Update 2016: This repo is not under active development, but I'm still happy to handle support requests.

Repo Contents

The repository has the following primary content:

contains relevant Matlab code. This should be the working dir in Matlab. within code/, you can find a fast intro script in code/demo/EasyDemo.m

This plain-text documentation file is your one-stop resource for getting started. Any further questions, please contact Mike.

contains one example dataset (6 Mocap sequences of various exercises) See the demos for how to run posterior inference on this data. Other example datasets (from our NIPS 2012 paper) are available by contacting Mike via email.

Academic Citation

If you find this toolbox useful, please cite one of our papers:

AOAS 2014 journal article

This long-form article presents a coherent reference for understanding the BP-HMM and BP-AR-HMM models, as well as an updated split-merge MCMC inference algorithm.

NIPS 2012 conference paper

Our NIPS 2012 paper introduced improved split-merge inference algorithm for the BP-HMM.

Acknowledgements

This code is inspired by (and heavily based upon) the BP-AR-HMM toolbox, released by Emily Fox. Most functions have been completely re-written for speed, readability, and extensibility, but Emily deserves most credit for coming up with the original solid inference algorithms.

I also thank