Lab Home | Phone | Search
Center for Nonlinear Studies  Center for Nonlinear Studies
 Home 
 People 
 Current 
 Affiliates 
 Visitors 
 Students 
 Research 
 ICAM-LANL 
 Publications 
 Conferences 
 Workshops 
 Sponsorship 
 Talks 
 Colloquia 
 Colloquia Archive 
 Seminars 
 Postdoc Seminars Archive 
 Quantum Lunch 
 Quantum Lunch Archive 
 CMS Colloquia 
 Q-Mat Seminars 
 Q-Mat Seminars Archive 
 P/T Colloquia 
 Archive 
 Kac Lectures 
 Kac Fellows 
 Dist. Quant. Lecture 
 Ulam Scholar 
 Colloquia 
 
 Jobs 
 Postdocs 
 CNLS Fellowship Application 
 Students 
 Student Program 
 Visitors 
 Description 
 Past Visitors 
 Services 
 General 
 
 History of CNLS 
 
 Maps, Directions 
 CNLS Office 
 T-Division 
 LANL 
 
Wednesday, June 23, 2010
3:00 PM - 4:00 PM
CNLS Conference Room (TA-3, Bldg 1690)

Seminar

Probabilistic Inference Using Divide & Concur and Belief Propagation

Jonathan Yedidia
MERL

In "probabilistic inference" problems one tries to estimate the true state of some quantities of interest, given only noisy or ambiguous sensor measurements of related quantities. Such problems arise in a very wide variety of fields, with applications in communications, signal processing, computer vision, speech and audio recognition, machine learning, physics, and robotics. A common formalism that can be used to attack all these problems is using message-passing algorithms that operate on "factor graphs."

In this talk, I will describe and compare two particularly important algorithms for probabilistic inference: the celebrated "belief propagation" (BP) algorithm; and the "divide and concur" (D&C) algorithm recently developed by Gravel and Elser, which I will show can also be understood as a message-passing algorithm. I will describe the relative advantages of the two algorithms, as well as various methods that can improve their performance. The D&C algorithm has some notable advantages compared with BP, in that it more naturally deals with problems with continuous valued variables, or with variables that lack local evidence. Another advantage of D&C is that its "difference-map" dynamics enables it to avoid "traps."

I will also describe a new decoder (developed with Yige Wang and Stark Draper) for low-density parity check codes that combines the ideas of the D&C and BP algorithms. This "difference-map belief propagation" (DMBP) decoder significantly improves error-floor performance compared with state-of-the-art BP decoders, while maintaining a similar computational complexity.

Host: Misha Chertkov