en
0.25
0.5
0.75
1.25
1.5
1.75
2
Beating Bandits in Gradually Evolving Worlds
Published on Aug 09, 20133206 Views
Consider the online convex optimization problem, in which a player has to choose actions iteratively and suffers corresponding losses according to some convex loss functions, and the goal is to minimi
Related categories
Chapter list
Beating Bandits in Gradually Evolving Worlds00:00
Outline00:06
Online Learning: Routing (Example) - 100:23
Online Learning: Routing (Example) - 200:34
Online Learning: Model - 100:43
Online Learning: Model - 201:53
Online Learning: Model - 302:19
Online Learning: Model - 402:29
Full Information Setting02:53
Online Problems02:59
Previous Results - 103:17
Previous Result: Deviation04:02
Previous Results - 204:37
Bandit Setting04:45
Previous Results - 304:49
Previous Results - 404:57
Previous Results - 505:12
Previous Results - 605:43
Difficulties in Bandit Setting05:53
Approach for Bandit - 105:58
Approach for Bandit - 206:23
Approach for Bandit - 306:43
Estimation - 107:23
Estimation - 208:26
Estimation - 308:33
Estimation - 408:37
Estimation - 508:48
Estimation - 609:30
Another Issue: Exploration09:48
Two-Point Bandit Setting11:07
Two-Point Bandit [ADX10]11:17
Motivation11:38
Previous Results - 711:59
Previous Results - 812:10
Our Results 12:30
Main Algorithm - 112:45
Gradient Descent12:59
Full information Algorithm’s idea - 113:43
Approach for Bandit14:40
Observation 15:12
First Try 16:03
Main Algorithm - 216:13
Full information Algorithm’s idea - 216:22
Algorithm’s Idea - 116:49
Algorithm - 117:42
Algorithm’s Idea - 218:36
Algorithm - 218:55
Analysis19:10
Results19:36
Thank you !!19:50