Bayesian offline RL can select hyperparameters and estimate deployed regret from posterior predictive uncertainty, with a regret bound at the parametric minimax rate.
Information-theoretic characterization of bayes performance and the choice of priors in parametric and nonparametric problems
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Fully Offline Reinforcement Learning
Bayesian offline RL can select hyperparameters and estimate deployed regret from posterior predictive uncertainty, with a regret bound at the parametric minimax rate.