Wednesday, October 31, 2001
Hewlett Packard Auditorium, 306 Soda Hall
4:00-5:00 p.m.

Dr. Robert Nowak

Electrical and Computer Engineering Department, Rice University

New ways to map the brain, the Internet and the cosmos: Image analysis by counting

Abstract:

This talk will describe statistical image analysis techniques used to create images of the brain, interpret high-energy cosmic events, and track performance characteristics throughout the Internet. In all three cases the acquisition of data involves counting (photons or bits/packets of information). This commonality is the basis for a
statistical approach to image analysis and reconstruction.

To illustrate the methodology, the "Internet Tomography" problem will be described in detail. Network performance information can be extremely useful, but for most applications it is important that the
information be localized. Knowing how portions or individual components of the network are performing is more valuable than
generating a global measure of performance. Network tomography is a cost effective process that produces accurate mappings of network structure and performance.

The nonpararmetric nature of all three imaging problems will be
examined, and ongoing work towards a unified framework for near
minimax-optimal estimation will be presented. Sequential inference
methods for dynamic, distributed networks (e.g., peer-to-peer wireless networks) can also be developed within this framework and ongoing work in this direction will be described.

Robert Nowak received the B.S. (with highest distinction), M.S., and Ph.D. degrees in electrical engineering from the University of Wisconsin-Madison in 1990, 1992, and 1995, respectively. He spent several summers with General Electric Medical Systems' Applied Science Laboratory, where he received the General Electric Genius of Invention Award and a U.S. patent for his work in 3-d computed tomography. He was an Assistant Professor at Michigan State University from 1996-1999. He is now an Assistant Professor at Rice University, Houston, Texas. His research interests include statistical image and signal processing, multiscale analysis, medical imaging, and communication networks. Dr. Nowak received the National Science Foundation CAREER Award in 1997, the Army Research Office Young
Investigator Program Award in 1999, the Office of Naval Research Young Investigator Program Award in 2000, and IEEE Signal Processing Society
Young Author Best Paper Award in 2000.