Here is the latest version of belief propagation for b-matching, which uses fast belief updates, handles function-based weights so you don't have to store the full weight matrix in memory and and includes a mex interface to the b-matching library. See the README for more info.


The old versions can be accessed below


Here is some new[er but old now] code to do bd-matching with different degree constraints on each node. New features include user set parameters for maximum iterations, damping. The new code operates on full graphs so it can be used as an approximation for bd-matching on unipartite graph. See the README for more info.



Here is the [old] code along with some documentation for b-matching. The code is in C, but I include a MATLAB version and a MATLAB interface for the C version. I also include pre-compiled Windows, Linux, and Mac OS X binaries.


Back to homepage.

Send Bert Huang an email if you have any questions.

Based on the paper (please cite it if you use this code):

Bert Huang and Tony Jebara, "Loopy Belief Propagation for Bipartite Maximum Weight b-Matching," in Proceedings of the Eleventh International Conference on Artificial Intelligence and Statistics , San Juan, Puerto Rico, 2007.

The code and documentation is intended for academic purposes. Note: please read this Columbia University license file if you intend to use this software for non-academic applications.