Tuesday, November 28, 2017 - 14:00 to 15:00
Hamilton Institute Seminar Room 317, 3rd Floor, Eolas Building, North Campus
Speaker: Professor David Leslie, Department of Mathematics and Statistics, Lancaster University, U.K.
Title: "Bandits defending borders"
Abstract: This talk will present work in the multi-armed bandit framework, which models sequential decision-making in which the decision-maker must both learn the value of actions and optimise at the same time. This is the dominant framework used to power online advertising and recommendation systems. The talk will cover work on Thompson sampling, a simple and probably consistent heuristic to solve the bandit problem in rather general settings as well as recent work in which we use a combinatorial bandit framework to model the defence of a border using multiple border defenders.