Regularization and Variance-Weighted Regression Achieves Minimax Optimality in Linear MDPs: Theory and Practice.
Toshinori KitamuraTadashi KozunoYunhao TangNino VieillardMichal ValkoWenhao YangJincheng MeiPierre MénardMohammad Gheshlaghi AzarRémi MunosOlivier PietquinMatthieu GeistCsaba SzepesváriWataru KumagaiYutaka MatsuoPublished in: CoRR (2023)
Keyphrases
- square loss
- aggregating algorithm
- markov decision processes
- simple linear
- reinforcement learning
- worst case
- optimal policy
- regression model
- linear regression
- reproducing kernel hilbert space
- dynamic programming
- locally linear
- model selection
- average cost
- average reward
- image restoration
- state space
- linear models
- initial state
- loss function
- variance reduction
- regression function
- kernel ridge regression