Publication Type

Conference Proceeding Article

Publication Date

5-2016

Abstract

Collective graphical models (CGMs) providea framework for reasoning about a populationof independent and identically distributed individualswhen only noisy and aggregate observationsare given. Previous approaches forinference in CGMs work on a junction-treerepresentation, thereby highly limiting theirscalability. To remedy this, we show how theBethe entropy approximation naturally arisesfor the inference problem in CGMs. We reformulatethe resulting optimization problem asa difference-of-convex functions program thatcan capture different types of CGM noisemodels. Using the concave-convex procedure,we then develop a scalable messagepassingalgorithm. Empirically, we showour approach is highly scalable and accuratefor large graphs, more than an orderof-magnitudefaster than a generic optimizationsolver, and is guaranteed to converge unlikethe previous message-passing approachNLBP that fails in several loopy graphs.

Discipline

Computer Sciences

Research Areas

Intelligent Systems and Decision Analytics

Publication

Proceedings of the 19th International Conference on Artificial Intelligence and Statistics (AISTATS) 2016, Cadiz, Spain

Volume

51

First Page

685

Last Page

693

Publisher

JMLR

City or Country

Cambridge, MA

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Additional URL

http://www.jmlr.org/proceedings/papers/v51/nguyen16b.pdf

Share

COinS