$α$ Belief Propagation as Fully Factorized Approximation

Belief propagation (BP) can do exact inference in loop-free graphs, but its\nperformance could be poor in graphs with loops, and the understanding of its\nsolution is limited. This work gives an interpretable belief propagation rule\nthat is actually minimization of a localized $\\alpha$-divergence. We term this\nalgorithm as $\\alpha$ belief propagation ($\\alpha$-BP). The performance of\n$\\alpha$-BP is tested in MAP (maximum a posterior) inference problems, where\n$\\alpha$-BP can outperform (loopy) BP by a significant margin even in\nfully-connected graphs.\n