Sign inSign up
Batched Bandit Problems thumbnail
Pause
Mute
Subtitles not available
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Batched Bandit Problems

Published on Aug 20, 20151733 Views

Philippe Rigollet

Motivated by practical applications, chiefly clinical trials, we study the regret achievable for stochastic multi-armed bandits under the constraint that the employed policy must split trials into a s

COLT 2015 - Paris

Related categories

Active LearningComputational Learning TheoryOn-line LearningReinforcement LearningSemi-supervised Learning

Chapter list

Batched Bandits00:00
Thompson (1933)00:11
Two armed bandit problem - 100:26
Two armed bandit problem - 200:33
Two armed bandit problem - 300:35
Two armed bandit problem - 400:37
Two armed bandit problem - 500:47
Two armed bandit problem - 600:49
Two armed bandit problem - 700:52
Two armed bandit problem - 800:53
Two armed bandit problem - 900:56
Two armed bandit problem - 1000:57
Two armed bandit problem - 1101:03
Two armed bandit problem - 1201:04
Comulative regret01:05
Explore-then-commit policy02:04
A clinical trial in NEJM02:56
Calculation03:21
Clinical trials03:33
Questions03:57
Batched bandits04:17
A generic explore-then-commit policy04:41
Stage sizes - 105:05
Stage sizes - 205:34
How it behaves06:02