Poster
A KL-LUCB algorithm for Large-Scale Crowdsourcing
Ervin Tanczos · Robert Nowak · Bob Mankoff
Pacific Ballroom #28
Keywords: [ Bandit Algorithms ]
This paper focuses on best-arm identification in multi-armed bandits with bounded rewards. We develop an algorithm that is a fusion of lil-UCB and KL-LUCB, offering the best qualities of the two algorithms in one method. This is achieved by proving a novel anytime confidence bound for the mean of bounded distributions, which is the analogue of the LIL-type bounds recently developed for sub-Gaussian distributions. We corroborate our theoretical results with numerical experiments based on the New Yorker Cartoon Caption Contest.
Live content is unavailable. Log in and register to view live content