Machine Learning Meetings and Events
Group Meetings: Group meetings are held Mondays from 11am to Noon (talk starts 11:10am) in D.L. Pratt 290C unless otherwise noted. Meetings are coordinated by Hugo Larochelle.
Tea Talks: Tea talks are held every Wednesday at 4:00pm in D.L. Pratt 290C. Talks should be simple, accessible, and not exceed 15 minutes. Speakers bring snacks, make tea, and provide a copy of the presented paper.
Group Meeting Sep 28, 2009: Max Product in High Order Factor Graphs
- Speaker: Danny Tarlow
- Abstract:
Graphical models are indispensable tools for representing the structure of local interactions between variables in probabilistic models. Unfortunately, inference becomes computationally expensive as the interactions in the network become less local. Even in the case of binary variables, standard approximate inference procedures such as loopy belief propagation quickly become intractable as the size of the largest clique grows.
In some cases, however, special structure in the potential allows for tractable MAP inference over high-order interactions. Some tractable structures are used in high order pattern-based clique potentials, max-product formulations of weighted b-matching, affinity propagation, and cluster size-based priors for exemplar-based clustering. We present a unifying view of these structures and generalize the class of factors that admit efficient max-product belief propagation inference.
We show several applications combining these factors in different ways to build intuitive high order models for correspondence, segmentation, and clustering tasks that would not be possible with only local interactions.