Abstract

In this talk I mainly focus on my work on methods and analysis of recursive algorithms for statistical estimation, classification, regression, and optimization under uncertainty. I also describe some significant applications of this work. I first discuss work on stochastic sampling and optimization, including the relationship between Markov chain and diffusion methods, and the formulation and analysis of a stochastic approximation for global optimization. Next, I discuss work on classification and regression trees, including iterative, incremental/adaptive, and non-Bayesian (Neyman Pearson) approaches. Also, stability and convergence analysis of the classical backpropagation algorithm for training multilayer neural networks is described. Next, I discuss work on several problems in low complexity adaptive linear estimation/filtering, including incorporating partial model information, analysis of variable and data-dependent step size, and fundamental limits on tracking time varying channels. I also briefly overview work on model-based techniques for signal and image processing, and problems in wireless and satellite communications and broadcast, the latter in collaboration with industry. Finally, I mention current work on time series analysis with application to the health sciences.

Biography

Saul B. Gelfand received the S.B. degree in Physics, and the Ph.D. degree in Electrical Engineering and Computer Science, both from the Massachusetts Institute of Technology, Cambridge, MA. Previously he was with Scientific Systems, Inc., Cambridge, MA, and Bolt Beranek and Newman Inc., Cambridge, MA. Since 1987 he has been with the School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, where he is Professor. His interests lie broadly in the modeling, analysis and optimization of stochastic signals and systems. His work has been supported by NSF, NIH, ARO, Thomson Multimedia, Motorola, Lucent Technologies and Northrop Grumman.

Watch Dr. Gelfand's Presentation