Artificial Intelligence


Exact Marginal Inference in General Markov Random Field Models Using Linear Programming

Authors: Ikhlef Bechar

This paper addresses the problem of exact marginal inference in general higher-order Markov random field (MRF) models. This is a fundamental AI problem, yet, renowned for its hardness. Nevertheless, by introducing an algebraic framework (referred to as the ortho-marginal framework)--which turns out to be, at once, a general approximation framework of discrete functions with arbitrary accuracy by means of their sets of margins, as well as a a principled means for modeling sets of locally consistent functions from a global perspective-- we are able to devise a linear programming approach which can solve the marginal inference problem for any instance of the MRF model both exactly and efficiently.

Comments: 36 Pages. added the omitted nonnegativity constraint on the pseudo-marginals in LP (16)

Download: PDF

Submission history

[v1] 2017-09-15 19:18:21
[v2] 2017-09-16 09:03:08
[v3] 2017-09-17 04:30:58

Unique-IP document downloads: 16 times is a pre-print repository rather than a journal. Articles hosted may not yet have been verified by peer-review and should be treated as preliminary. In particular, anything that appears to include financial or legal advice or proposed medical treatments should be treated with due caution. will not be responsible for any consequences of actions that result from any form of use of any documents on this website.

Add your own feedback and questions here:
You are equally welcome to be positive or negative about any paper but please be polite. If you are being critical you must mention at least one specific error, otherwise your comment will be deleted as unhelpful.

comments powered by Disqus