Lab Home | Phone | Search
Center for Nonlinear Studies  Center for Nonlinear Studies
 Home 
 People 
 Current 
 Affiliates 
 Alumni 
 Visitors 
 Students 
 Research 
 ICAM-LANL 
 Quantum 
 Publications 
 Publications 
 2007 
 2006 
 2005 
 2004 
 2003 
 2002 
 2001 
 2000 
 <1999 
 Conferences 
 Workshops 
 Sponsorship 
 Talks 
 Colloquia 
 Colloquia Archive 
 Seminars 
 Postdoc Seminars Archive 
 Quantum Lunch 
 CMS Colloquia 
 Q-Mat Seminars 
 Q-Mat Seminars Archive 
 Archive 
 Kac Lectures 
 Dist. Quant. Lecture 
 Ulam Scholar 
 Colloquia 
 
 Jobs 
 Students 
 Summer Research 
 Student Application 
 Visitors 
 Description 
 Past Visitors 
 Services 
 General 
 PD Travel Request 
 
 History of CNLS 
 
 Maps, Directions 
 CNLS Office 
 T-Division 
 LANL 
 
Monday, December 11, 2006
3:00 PM - 4:00 PM
CNLS Conference Room

Colloquium

Graph-Cover Decoding: Connecting Iterative Decoding and Linear Programming Decoding

Pascal Vontobel
Hewlett-Packard Laboratories

Whenever information is transmitted across a channel, we have to ensure its integrity against errors. The ground-breaking work of Shannon showed (at least theoretically) how such integrity can be achieved, namely by using an appropriately chosen encoder at the sender side and an appropriately chosen decoder at the receiver side.

From a practical point of view, so-called low-density parity-check (LDPC) and turbo codes together with the class of message-passing iterative decoders have become increasingly popular in the last decade. Another interesting class of decoders is the class of linear programming decoders that was recently introduced by Feldman, Karger, and Wainwright.

In this talk we discuss the so-called graph cover decoder, a theoretical tool that shows why the behavior of these two above-mentioned classes of decoders are tightly connected. These connections allow one to characterize a given finite-length code under iterative decoding, in particular, one can study a (possibly existing) error floor in the word error-rate vs. signal-to-noise-ratio curve. Let us emphasize that this is in stark contrast to most other existing iterative decoding analysis techniques which only characterize ensembles of codes or infinitely long codes.

(This talk is based on joint work with Ralf Koetter, UIUC, and is planned to be accessible to an audience with a basic knowledge in probability theory and optimization theory.)


Biography: Pascal O. Vontobel received a diploma in electrical engineering in 1997, a post-diploma in information techniques in 2002, and a PhD degree in electrical engineering in 2003, all from ETH Zurich, Switzerland. After being a postdoctoral research associate at the University of Illinois at Urbana-Champaign, the University of Wisconsin-Madison (visiting assistant professor), and at the Massachusetts Institute of Technology, he joined the Information Theory Research Group at Hewlett-Packard Laboratories in Palo Alto, CA, in the summer of 2006 as a research scientist. For his PhD thesis he was awarded the ETH medal.

Dr. Vontobel is interested in information theory and signal processing in general. More specifically, for his diploma thesis he worked on source coding. Since then, he has mainly looked at the construction of LDPC and turbo codes based on algebraic principles, the calculation and bounding of capacities and information rates of finite-state machine channels, and connections between factor graphs, the summary-product algorithm, and electrical networks. Most recently, he has worked towards an understanding and characterization of the summary-product algorithm on factor graphs with cycles and its connections to linear programming (LP) decoding.

Host: Misha Chertkov