Mehryar Mohri IACS Talk
At this weeks IACS seminar series, Mehryar Mohri (NYU/Google) will discuss how weighted automata can be used for online learning.
The talk will take place Thursday, September 13, at 1pm in the IACS seminar room.
Title: Competing with Weighted Automata Experts
Abstract: This talk discusses a general framework for online learning with expert advice where regret is defined with respect to a dynamic set of experts represented by a weighted automata. This framework is highly relevant to applications and covers many previous studies.
We present a series of algorithms for this problem including an automata-based algorithm extending weighted-majority. We also give a more efficient algorithm drawing from automata theory tools and algorithms. We further present efficient algorithms based on an approximation of the competitor weighted automaton, in particular , n-gram models obtained by minimizing the infinite-Renyi divergence, and present an extensive study of the approximation properties of such models. Finally, we also extend our algorithms and results to the framework of sleeping experts. Finally, we show that our approximation methods can be extended to online convex optimization and a general mirror descent setting.