AAAI Publications, Twenty-Sixth AAAI Conference on Artificial Intelligence

Font Size: 
Approximating the Sum Operation for Marginal-MAP Inference
Qiang Cheng, Feng Chen, Jianwu Dong, Wenli Xu, Alexander Ihler

Last modified: 2012-07-14

Abstract


We study the marginal-MAP problem on graphical models, and present a novel approximation method based on direct approximation of the sum operation. A primary difficulty of marginal-MAP problems lies in the non-commutativity of the sum and max operations, so that even in highly structured models, marginalization may produce a densely connected graph over the variables to be maximized, resulting in an intractable potential function with exponential size. We propose a chain decomposition approach for summing over the marginalized variables, in which we produce a structured approximation to the MAP component of the problem consisting of only pairwise potentials. We show that this approach is equivalent to the maximization of a specific variational free energy, and it provides an upper bound of the optimal probability. Finally, experimental results demonstrate that our method performs favorably compared to previous methods.

Full Text: PDF