Room 722, Institute of Mathematics, Academia Sinica
(中研院數學所 722室)
Large deviation theory with connections to statistical physics and information theory I & II
Lo-Bin Chang (Ohio State University)
Abstract:
Large deviation theory appears to be a powerful mathematical tool in applied probability when analyzing the probabilities of extreme or tail events. The basic formulation of the theorems is based on the “rate function,” which is highly related to Shannon Entropy in information theory and Gibbs distributions in statistical mechanics. In these two talks, I will introduce the large deviation principle and discuss its connections to statistical physics and information theory. Topics covered include Gibbs ensembles, maximum entropy principle, (relative) entropy and lossless source coding.