Date: December 2, 2005

Time: 3:00 p.m.

Location: Centergy One, Room 5186

Speaker(s): Soner Ozgur

Title: Performance Analysis of Particle Filter Proposal Functions

Abstract:
Recently, Monte Carlo methods have been applied to various fields in signal processing. Sequential Monte Carlo (SMC) methods have received special attention because of their ability to track dynamic systems. It has also been found that the convergence properties of SMC algorithms depend on the choice of the proposal distribution. In this work, we propose a method to analyze and quantify the convergence rates of SMC algorithms. The approach makes it possible to quantify the effectiveness of the candidate proposal distributions.

Bio:
Soner Ozgur received his BS in Electrical Engineering from USC in 1998. He received his MS in ECE from GaTech in 2000. Since 2000, he has been in CSIP. His research interests include multiuser detection, MIMO, equalization, iterative/turbo receivers and, more recently, Monte Carlo algorithms for telecom applications.

Date:	December 2, 2005
Time:	3:00 p.m.
Location:	Centergy One, Room 5186
Speaker(s):	Soner Ozgur
Title:	Performance Analysis of Particle Filter Proposal Functions
Abstract:	Recently, Monte Carlo methods have been applied to various fields in signal processing. Sequential Monte Carlo (SMC) methods have received special attention because of their ability to track dynamic systems. It has also been found that the convergence properties of SMC algorithms depend on the choice of the proposal distribution. In this work, we propose a method to analyze and quantify the convergence rates of SMC algorithms. The approach makes it possible to quantify the effectiveness of the candidate proposal distributions.
Bio:	Soner Ozgur received his BS in Electrical Engineering from USC in 1998. He received his MS in ECE from GaTech in 2000. Since 2000, he has been in CSIP. His research interests include multiuser detection, MIMO, equalization, iterative/turbo receivers and, more recently, Monte Carlo algorithms for telecom applications.