Selecting the state representation in reinforcement Learning thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Selecting the state representation in reinforcement Learning

Published on Jan 25, 20124021 Views

The problem of selecting the right state-representation in a reinforcement learning problem is considered. Several models (functions mapping past observations to a finite set) of the observations are

Related categories

Chapter list

Selecting the state representation in reinforcement Learning00:00
States in MDP - 0100:24
Introduction01:23
State representation01:36
Challenge02:07
Some Motivations03:26
Some Motivations (II)04:09
Example: high-level feature selection04:42
States in MDP - 0204:45
Regret05:34
UCRL2 as a subroutine algorithm05:48
The Best Lower Bound algorithm - 0107:24
The Best Lower Bound algorithm - 0208:36
The Best Lower Bound algorithm - 0309:35
The Best Lower Bound algorithm - 0409:55
The Best Lower Bound algorithm - 0510:04
The Best Lower Bound algorithm - 0610:06
Intuition of the proposed algorithm (I)10:40
Intuition of the algorithm (II)11:10
Result11:35
Discussion15:43
Future work, open questions17:01