ITR Seminar Announcement
| Presenter(s) | Dr Pascal Vontobel |
| Title | A Connection Between Channel Coding Linear
Programming Decoding and
Compressed Sensing Linear Programming Decoding
|
| Date | Wednesday 11 Nov 2009 |
| Time | 13:00 |
| Location | SPRI Lecture Theatre (1-6) |
Seminar Abstract
Recently there has been substantial interest in the
theory of recovering
sparse approximations of signals that satisfy
linear measurements. Compressed
(or compressive) sensing research [Donoho,
Candes/Tao, etc.] has developed
conditions for measurement matrices under which
(approximately) sparse signals
can be recovered by solving a linear programming
relaxation of the original
NP-hard combinatorial problem. Interestingly, in
one of the first papers in
this area, Candes and Tao presented a setup they
called "decoding by linear
programming," henceforth called CS-LPD, where the
(approximately) sparse
signal corresponds to real-valued noise that is
added to a real-valued signal
that is to be recovered in a hypothetical
communication problem.
At about the same time, in an independent line of
research, Feldman,
Wainwright, and Karger considered the problem of
decoding a binary linear code
that is used for data communication over a binary-
input memoryless channel, a
problem that is also NP-hard in general. Feldman et
al. formulated this
channel coding problem as an integer linear
program, along with presenting a
linear programming relaxation for it, henceforth
called CC-LPD.
CS-LPD and CC-LPD (and the setups they are derived
from) are *formally* very
similar, however, it is rather unclear if there is
a connection beyond this
formal relationship. In fact Candes and Tao in
their original paper asked the
following question: ``... In summary, there does
not seem to be any explicit
known connection with that line of work (i.e.,
Feldman et al.'s work) but it
would perhaps be of future interest to explore if
there is one.''
In this talk we present such a connection between
CS-LPD and CC-LPD. The
general form of our results is that if a given
binary parity-check matrix is
"good" for CC-LPD then the same matrix (considered
over the reals) is a "good"
measurement matrix for CS-LPD. The notion of a
"good"
Presenter Biography
Pascal O. Vontobel received the Diploma degree in
electrical engineering in
1997, the Post-Diploma degree in information
techniques in 2002, and the
Ph.D. degree in electrical engineering in 2003, all
from ETH Zurich, Zurich,
Switzerland.
From 1997 to 2002, he was a Research and Teaching
Assistant at the Signal and
Information Processing Laboratory at ETH Zurich.
After being a Postdoctoral
Research Associate at the University of Illinois at
Urbana-Champaign, at 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. His research
interests lie in information
theory, communications, and signal processing.
He has been on the technical program committees of
several international
conferences, he has recently co-organized a BIRS
workshop in Banff on
"Applications of
Back to seminar list