Subrahmanian delivers distinguished lecture at UAEU

ISR-affiliated Professor and Director of the University of Maryland Institute for Advanced Computer Studies (UMIACS) V.S. Subrahmanian (CS/UMIACS) presented a distinguished lecture at United Arab Emirates University on Feb. 23. Subrahmanian was invited by former ISR Director Eyad Abed (ECE/ISR), dean of UAEU’s College of Information Technology.

Subrahmanian spoke about his research in Mining Stochastic Models for Forecasting. In this work, Subrahmanian uses modeling to better understand and predict various phenomena. For instance; public health organizations might want to model the behavior of certain diseases, while investors might want to model behaviors of large institutional investors. In the lecture, Subrahmanian explained his “Stochastic Opponent Modeling Agents” (SOMA). Given a subject to model, and a body of historical data pertinent to that subject, SOMA can easily learn models of behaviors of the subject. These models are represented as stochastic rules of the form "If some condition is true on certain variables, then the subject will exhibit behavior B with some probability." Once such a model is derived, Subrahmanian and his team develop algorithmic methods to forecast what the subject will do in a given situation (real or hypothetical).

Related Articles:
Davis, Yu, Subrahmanian finalists at Invention of the Year Awards
SCARE software predicts locations from observations and known constraints
Subrahmanian's CAGE software platform recreates complex real-life situations for simulations

March 8, 2010