Scaling MPE Inference for Constrained Continuous Markov Random Fields with Consensus Optimization

DMCA

Scaling MPE Inference for Constrained Continuous Markov Random Fields with Consensus Optimization

Cached

Download Links

by Stephen H. Bach , Lise Getoor , Matthias Broecheler , Dianne P. O’leary

Citations:

17 - 14 self

BibTeX

@MISC{Bach_scalingmpe,
    author = {Stephen H. Bach and Lise Getoor and Matthias Broecheler and Dianne P. O’leary},
    title = {Scaling MPE Inference for Constrained Continuous Markov Random Fields with Consensus Optimization},
    year = {}
}

OpenURL

Abstract

Probabilistic graphical models are powerful tools for analyzing constrained, continuous domains. However, finding most-probable explanations (MPEs) in these models can be computationally expensive. In this paper, we improve the scalability of MPE inference in a class of graphical models with piecewise-linear and piecewise-quadratic dependencies and linear constraints over continuous domains. We derive algorithms based on a consensus-optimization framework and demonstrate their superior performance over state of the art. We show empirically that in a large-scale voter-preference modeling problem our algorithms scale linearly in the number of dependencies and constraints. 1

Keyphrases

mpe inference consensus optimization constrained continuous markov random field continuous domain consensus-optimization framework algorithm scale superior performance linear constraint most-probable explanation probabilistic graphical model large-scale voter-preference modeling problem graphical model powerful tool piecewise-quadratic dependency

DMCA