Minimax-Bayes Reinforcement Learning
Author(s)
Thomas Kleine Buening
Hannes Eriksson
Divya Grover
Emilio Jorge
Date issued
2023
In
PMLR
Vol
206
Abstract
While the Bayesian decision-theoretic framework offers an elegant solution to the problem of decision making under uncertainty, one question is how to appropriately select the prior distribution. One idea is to employ a worst-case prior. However, this is not as easy to specify in sequential decision making as in simple statistical estimation problems. This paper studies (sometimes approximate) minimax-Bayes solutions for various reinforcement learning problems to gain insights into the properties of the corresponding priors and policies. We find that while the worst-case prior depends on the setting, the corresponding minimax policies are more robust than those that assume a standard (i.e. uniform) prior.
Event name
AISTATS
Location
Valencia, Spain
Publication type
conference proceedings
File(s)![Thumbnail Image]()
Loading...
Name
buening23a.pdf
Type
Main Article
Size
1.11 MB
Format
Adobe PDF
Checksum
(MD5):4ac208e75cc59505b82b533a53462c36