Capacity of Interference Networks with a Local View

Date: 
Thursday, May 14, 2009 - 3:36pm

UCSB COMPUTER SCIENCE DEPARTMENT PRESENTS:
THURSDAY, MAY 28, 2009
2:00 – 3:00
Computer Science Conference Room, Harold Frank Hall Rm. 1132

HOST: Heather Zheng

SPEAKER: ASHU SABHARWAL
Assistant Professor, Electrical and Computer Engineering
Rice University

Title: Capacity of Interference Networks with a Local View

Abstract:

A key element of mobile wireless networks is their distributed nature,
which implies that often nodes only have a local view of the network.
This leads to a situation where the ground truth for each node is
different from other nodes. Yet, they have to make decisions about their
transmission parameters like rate and power, such that their collective
decision does not violate network capacity limits but still maximize
spectral efficiency.

In this talk, we will first formulate a message-passing protocol which
allows the information about the network to trickle via local message
forwarding. The protocol naturally gives rise to networks where nodes
have different amount of local information. We will then propose a
distributed rate-allocation policy and analyze its performance in some
worst case topologies. The analysis systematically captures the extent
of loss in network capacity which is incurred when nodes make decisions
based on a local view of the network.

Bio:

Ashutosh Sabharwal received his B.Tech from Indian Institute of
Technology, New Delhi in 1993. He graduated from The Ohio State
University with a MS in 1995 and PhD in 1999. Currently he is an
Assistant Professor at Rice, where he also the Director of Center for
Multimedia Communication. His main research interests are information
theoretic foundations, protocols and platforms for high performance
wireless networks. He is the founder of WARP project.