For robust average-reward MDPs, the paper proves model-free Q-learning and actor-critic algorithms converge with tilde O(epsilon^{-2}) sample complexity via a carefully constructed semi-norm contraction.
Distributionally robust stochastic optimization with W asserstein distance
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Efficient Q-Learning and Actor-Critic Methods for Robust Average-Reward Reinforcement Learning
For robust average-reward MDPs, the paper proves model-free Q-learning and actor-critic algorithms converge with tilde O(epsilon^{-2}) sample complexity via a carefully constructed semi-norm contraction.