|
Vibhav Giridhar Gogate |
|
Research Overview |
|
My research interests are in the automated reasoning field of Artificial Intelligence focusing on graphical models such as Bayesian, constraint and Markov networks, influence diagrams and Markov decision processes, which compactly describe probabilistic and/or deterministic dependencies among a set of random variables. My goal is to develop practical algorithms that derive solutions to various reasoning tasks (or queries) defined on a graphical model. Because graphical models have tremendous applications in numerous fields like speech recognition, computer vision, biology, astronomy and hardware/software verification, my research has potential ramifications in many fields. It is well-known that most reasoning tasks defined over a graphical model are intractable in theory, typically #P-complete (e.g., model counting, partition function and marginal computation) or PSPACE-complete (e.g. Maximum a Posteriori or MAP problem). Obviously, unless P=NP, there does not exist a polynomial time algorithm for solving them and approximate techniques are a practical necessity. Therefore progress in approximate inference techniques would play a pivotal role in advancing the practical applicability of graphical models. Hence, in my PhD. thesis, I developed a range of approximate inference algorithms that utilize insights derived from graph theory, probability theory, statistics, statistical physics and logical reasoning. For more information, see my CV and publications list. |
|
PhD. candidate Donald Bren School of Information and Computer Sciences
Phone: (949)232-6363 (Cell) Office: DBH 4099 Email: vgogate at ics dot uci dot edu
Advisor: Rina Dechter |