Speaker: Ma Hui

Beijing Computational Science Research Center    

Abstract: Important inference problems in statistical physics, computer vision, error-correcting coding theory, and artificial intelligence can all be reformulated as the computation of marginal probabilities on factor graphs. In the presentation, the widely-adopted message-passing algorithm Belief Propagation would be introduced, which is an efficient way to solve the problem. And, we would explain the equivalency of Belief Propagation algorithm and the Bethe approximation of the free energy in mean-field statistical physics. Additionally, we would show the problems we met at implementing the algorithm and the approximation.

Date&Time: February 1, 2013 (Friday), 14:30–15:30
Location: 606 Conference Room