Talks & Seminars
Title: Approximate Methods for Probabilistic Inference in Dynamic Bayesian Networks
Dr. S. Akshay, National University of Singapore
Date & Time: August 11, 2011 15:40
Venue: Conference Room, C Block, First Floor, Kanwal Rekhi Building
Probabilistic models are often used to describe the dynamics of biochemical networks. To analyze these models one must compute the probability of a state at a given time in a large Markov chain, which can then be used to develop quantitative model checking techniques. However, doing this exactly is intractable for large networks and hence succinct representations and approximate methods of computation are needed. We consider the setting where we exploit the structure of the Markov chain to represent it succinctly as a Dynamic Bayesian Network (DBN). Now, exact analysis corresponds to performing probabilistic inference on the DBN, which is still infeasible in general. In this light, we will describe an algorithm that performs this computation approximately. We analyse the error made by the algorithm and come up with new variants where we trade-off computational speed for accuracy.
Speaker Profile:
List of Talks


Faculty CSE IT
Forgot Password
    [+] Sitemap     Feedback