BErMin: A Model Selection Algorithm for Reinforcement Learning Problems thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

BErMin: A Model Selection Algorithm for Reinforcement Learning Problems

Published on Jan 25, 20124077 Views

We consider the problem of model selection in the batch (offline, non-interactive) reinforcement learning setting when the goal is to find an action-value function with the smallest Bellman error amon

Related categories

Chapter list

BErMin: A Model Selection Algorithm for Reinforcement Learning Problems00:00
Given some interaction data - 0100:33
Given some interaction data - 0201:12
How to choose the architecture of the function approximator?01:41
Solution: Adaptive Algorithms02:35
Problem Setup - 0103:45
Problem Setup - 0203:51
Problem Setup - 0305:44
Challenge07:38
What about estimating the effect - 0109:14
What about estimating the effect - 0209:32
Not done yet! - 0111:21
Not done yet! - 0211:33
Algorithm 1 - 0113:44
Algorithm 1 - 0213:52
Assumptions16:19
Theorem16:44
Remember - 0117:19
Remember - 0217:22
Remember - 0317:29
Conclusion18:30