ECE Departmental Seminar
Signal processing in the Extreme
Dr. Douglas E. Johnston
Famingdale State College
Friday, 9/14/18, 11:00am
Light Engineering 250
Abstract: In this talk, we illustrate a new approach for analyzing extreme values such as large losses in financial markets (i.e., “market crashes”). We apply a stochastic parametrization for a generalized extreme value distribution to model the asymptotic behavior of the block-maxima for the time-series of interest. This allows for a relaxation of the iid assumption and the use of finer data blocks. By using a particle filter, with Rao-Blackwellization, we reduce the parameter space and provide a novel, recursive solution. We use this to compute the conditional predictive distribution that is used to assess the probability of extreme events and risk-measures. We introduce a new risk-measure, p-VaR, that is a more robust estimate of the true nature of value-at-risk, and illustrate our results using both simulated and stock market data from 1928-2017.
Bio: Dr. Johnston is an assistant professor in the applied mathematics department at Farmingdale State College and received his Ph.D. (1994) in electrical engineering from Stony Brook University. Prior to joining academia in 2017, Dr. Johnston worked on Wall Street, since 1994, in research and proprietary trading at a hedge fund and an investment bank. He has also been an adjunct professor at Stony Brook University in both the engineering and business schools. Prior to joining Wall Street, Dr. Johnston was a research engineer in the defense industry. He is a member of the IEEE Signal Processing Society and the Mathematical Association of America and has written numerous articles in the area of financial signal processing. His research interests are in extreme value theory, quantitative risk-management, and high-dimensional dynamical systems & signal processing.