Beyond Confidence Regions: Tight Bayesian Ambiguity Sets for Robust MDPs.

Marek Petrik Reazul Hasan Russel

Published in: CoRR (2019)

Keyphrases

lower bound
markov decision processes
reinforcement learning
upper bound
worst case
image regions
high confidence
optimal policy
keypoints
bayesian networks
maximum likelihood
parameter space
reward function
np hard
confidence level
context specific
decision theoretic planning
factored mdps