Seminars/Colloquia/Invited Talks

Seminars

Brandon Westover

Pattern Recognition with Budget Constraints

PLACE:	Clark 314
EVENT:	CIS Seminar Series
DATE:	December 6, 2005
TIME:	1:00 - 2:00

Abstract

How much information is needed for reliable pattern recognition? In other words:
1) What is the minimum amount of information an animal or machine can store in memory and still reliably distinguish the important patterns in its sensory data?
2) How much information must a pattern recognition system extract from incoming sensory signals?
3) Given fixed budgets for memory and sensory data representation, in how complex an environment may a recognition system conceivably operate?
Dr. Westover will discuss how to formalize these questions within an information theoretic framework, and present progress to date towards solving the general case and two important special cases.

Brief biography

Brandon Westover is a student in his last year of MD/PhD training at Washington University in Saint Louis. His research involves computational modeling of the primate visual system, and information theoretic aspects of statistical pattern recognition.