HaMMLET is a fast Forward-Backward Gibbs sampler for Bayesian inference on Hidden Markov Models (HMM). It uses the Haar wavelet transform to dynamically compress the data based on the current variance sample in each iteration. This yields speedup of up to two orders of magnitude on array CGH data, as well as vastly improved convergence characteristics. For details, please refer to the following preprint
John Wiedenhoeft, Eric Brugel, Alexander Schliep: "Fast Bayesian Inference of Copy Number Variants using Hidden Markov Models with Wavelet Compression". bioRxiv (2015). URL: http://biorxiv.org/content/early/2015/07/31/023705. DOI: 10.1101/023705 .
The version of HaMMLET used in the paper has been released; please click "View on GitHub" and checkout the "biorxiv" branch. As this release will remain unchanged for reproducibility purposes, we do not recommend using it in a production environment. For the latest stable release, please checkout the "master" branch.