Cooperative Graphical Models

Cooperative Graphical Models

Abstract

We study a rich family of distributions that capture variable interactions signifi-cantly more expressive than those representable with low-treewidth or pairwisegraphical models, or log-supermodular models. We call thesecooperative graph-ical models. Yet, this family retains structure, which we carefully exploit forefficient inference techniques. Our algorithms combine the polyhedral structure ofsubmodular functions in new ways with variational inference methods to obtainboth lower and upper bounds on the partition function. While ourfully convexupperbound is minimized as an SDP or via tree-reweighted belief propagation, our lowerbound is tightened via belief propagation or mean-field algorithms. The resultingalgorithms are easy to implement and, as our experiments show, effectively obtaingood bounds and marginals for synthetic and real-world examples.

Grafik Top
Authors
  • Djolonga, Josip
  • Jegelka, Stefanie
  • Tschiatschek, Sebastian
  • Krause, Andreas
Grafik Top
Shortfacts
Category
Paper in Conference Proceedings or in Workshop Proceedings (Paper)
Event Title
Neural Information Processing Systems (NIPS)
Divisions
Data Mining and Machine Learning
Event Location
Barcelona, Spain
Event Type
Conference
Event Dates
05.-10.12.2016
Series Name
Advances in Neural Information Processing Systems 29 (NIPS 2016)
Date
5 December 2016
Official URL
https://papers.nips.cc/paper/6122-cooperative-grap...
Export
Grafik Top