Belief Propagation module for Multi-Core architecture released

The module provides C++ code to easily create a Graphical Model and run either the Junction Tree algorithm (exact inference), or Loopy Belief Propagation (approximate inference). The code uses OpenMP for efficient processing on a multi-core architecture.

The Belief Propagation module can be download by executing

svn co http://elefant-svn.developer.nicta.com.au/elefant/trunk/elefant/beliefpropc

Coming soon: The author of the code, Julian McAuley, has recently published new research results which allow to accelerate computations for all Graphical Models in which the order of the clique potentials is lower than the order of the maximal clique (Exact Inference in Graphical Models: is There More to it?)

Please watch this space for future improvements in the code utilising theses recent research results.