Batched Bandit Problems
Copyright policy )Batched Bandit Problems
Motivated by practical applications, chiefly clinical trials, we study the regret achievable for stochastic bandits under the constraint that the employed policy must split trials into a small number of batches. We propose a simple policy, and show that a very small number of batches gives close to minimax optimal regret bounds. As a byproduct, we derive optimal policies with low switching cost for stochastic bandits.
Published at http://dx.doi.org/10.1214/15-AOS1381 in the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)
- University of Paris France
- National Bureau of Economic Research United States
- Princeton University
- New York University United States
- University of Paris France
multi-phase allocation, 62C20, switching cost, Minimax procedures in statistical decision theory, Mathematics - Statistics Theory, Statistics Theory (math.ST), 510, Applications of statistics to biology and medical sciences; meta analysis, grouped clinical trials, Multi-armed bandit problems, Sequential statistical design, 62L05, batches, sample size determination, multi-armed bandit problems, FOS: Mathematics, regret bounds
multi-phase allocation, 62C20, switching cost, Minimax procedures in statistical decision theory, Mathematics - Statistics Theory, Statistics Theory (math.ST), 510, Applications of statistics to biology and medical sciences; meta analysis, grouped clinical trials, Multi-armed bandit problems, Sequential statistical design, 62L05, batches, sample size determination, multi-armed bandit problems, FOS: Mathematics, regret bounds
selected citations These citations are derived from selected sources.This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).56 popularity This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.Top 1% influence This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).Top 10% impulse This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.Top 10%
Citations provided by BIP!Popularity provided by BIP!