Please join us this Friday, October 25th at 2:00 pm in the Library’s Hatfield Room for a presentation by Haiyan Cheng (Assistant Professor of Computer Science).
Title: How to Forecast What We Can’t See
Abstract: In chess, each player sees the positions and moves of his or her opponent. In card games, each player sees his or her cards, and possibly the used cards, but does not see unused cards of the opponent. We say that chess is a fully observable system, and card games are partially observable systems. There are many partially observable systems in nature. For example, in a weather system, we can measure specific quantities at specific time and location, such as temperature, wind speed, but the continuous true system state is unobservable due to our incomplete knowledge about the nature and the errors in modeling process. In a stock market, the stock prices are observable, but investor beliefs, which help drive prices, are not observable. In target tracking applications, the data collected through sensors is observable, but the dynamic position of the target is unobservable.
Professof Cheng will talk about the application of mathematical and statistical techniques to the estimation and forecasting of the unobservable part from the observable part, when both parts evolve randomly over time, and will report on her research in this area.