LECTURE: Non-asymptotic Information Theory

Speaker: Sergio Verdú

Organization: CEIT

Non-Asymptotic Information Theory

Day and Time: July 6, 2011 (11:00 – 12:30h)

Biography:

Sergio Verdú is the Eugene Higgins Professor of Electrical Engineering at Princeton University. A member of the National Academy of Engineering, Verdu is the recipient of the 2007 Claude E. Shannon Award and the 2008 IEEE Richard W. Hamming Medal. He was awarded a Doctorate Honoris Causa from the Universitat Politecnica de Catalunya in 2005. His research has received several awards including the 1998 Information Theory Outstanding Paper Award, the Information Theory Golden Jubilee Paper Award, and the 2006 Joint Communications/Information Theory Paper Award.

Abstract:

In real-time voice and high speed data applications, limited delay is a key design constraint; indeed, packet sizes as short as a few hundred bits are common in wireless systems. Traditional results on the fundamental limits of data compression and data transmission through noisy channels apply to the asymptotic regime as delay (or block length) goes to infinity.

In this talk, the speaker will review recent progress on the analysis of the fundamental limits as a function of blocklength. Going beyond traditional refinements to the fundamental asymptotic information theoretic limits, he will investigate the backoff from capacity (in channel coding) and the overhead over entropy (in lossless compression) and the rate-distortion function (in lossy source coding) incurred by coding at a given blocklength.

Requiring new proof techniques transcending traditional ones, our approach has dual components:  computable upper/lower bounds tight enough to reduce the uncertainty on the non-asymptotic fundamental limit to a level that is negligible compared to the gap to the long-blocklength asymptotics; and analytical approximations to the bounds that are accurate even for short blocklengths.

Sponsored by the Consolider Project «Foundations and Methodologies of Future Communication and Sensor Networks (COMONSENS)» Institutions: UValencia, UPC, UPF, UPM, UC3M, UC, UDC, UVigo, US, UNavarra.

Twitter Digg Delicious Stumbleupon Technorati Facebook Email

Los comentarios están cerrados.