Fall 1998 Operations Research Seminar Series
Joint Seminar with Statistics/Biomathematics
North Carolina State University

presents

Dr. James C. Spall

The Johns Hopkins University, Applied Physics Laboratory

"The Simultaneous Perturbation Method for System Optimization"

ABSTRACT

Multivariate optimization algorithms play a major role in virtually all areas of engineering and the physical and social sciences. There has recently been much interest in recursive optimization algorithms that rely on measurements of only the objective function to be optimized, not requiring direct measurements of the gradient of the objective function. Such algorithms have the advantage of not requiring detailed modeling information describing the relationship between the parameters to be optimized and the objective function. For example, many systems involving human beings are difficult to model, and could potentially benefit from such an optimization approach. In this spirit, the speaker will discuss the "simultaneous perturbation stochastic approximation (SPSA)" algorithm for difficult multivariate optimization problems arising in stochastic systems. SPSA has recently attracted considerable international attention in areas such as statistical parameter estimation, pattern recognition, feedback control, simulation-based optimization, and experimental design. The essential feature of SPSA--which accounts for its power and relative ease of use in difficult multivariate optimization problems--is the underlying gradient approximation that requires only two objective function measurements regardless of the dimension of the optimization problem. This talk will focus on the basic ideas and motivation behind SPSA without dwelling on the mathematical details. Examples will be drawn from problems in process control, sensor placement, and simulation-based optimization. according to the time available and interests of the audience.

Dr. James C. Spall

Dr. James C. Spall was the top graduate in School of Engineering, Oakland University, where he received his B.S. in systems engineering. He obtained his S.M. in Technology and Policy from MIT and his Ph.D in Systems Engineering from the University of Virginia in 1983. Dr. Spall joined the John Hopkins University's Applied Physics Lab in 1983 and was appointed to the Principal Professional Staff in 1991. He also teaches in the JHU School of Engineering. His work is concentrated at the interface of statistics and systems engineering, and he has served as project leader for many defense and civilian projects in these areas. He received the R.W. Hart Prize as principal investigator of the most outstanding Independent Research and Development project at JHU/APL for the year 1990. Dr. Spall has published numerous research papers in statistics and control, including articles on subjects such as time series, optimization, small-sample data analysis, parameter estimation, adaptive control, and neural networks. He holds two patents for inventions related to control systems. He is an Associate Editor for the IEEE Transactions on Automatic Control, a contributing editor for the Current Index to Statistics, and served as editor and co-author of the book Bayesian Analysis of Time Series and Dynamic Models (1988). He has a book forthcoming entitled Introduction to Stochastic Search and Optimization (Wiley). Dr. Spall has given many invited talks at national and international professional meetings. He is a senior member of IEEE, a member of the American Statistical Association and Sigma Xi, and a Fellow of the engineering honor society Tau Beta Pi.

Tuesday, October 27, 1998

4:00 pm

Riddick 320

Soft drinks, coffee and cookies will be available in this same room from 3:50 pm onwards.