Regret Bounds for Risk-Sensitive Reinforcement Learning.
Osbert BastaniYecheng Jason MaEstelle ShenWanqiao XuPublished in: CoRR (2022)
Keyphrases
- risk sensitive
- reinforcement learning
- model free
- optimal control
- markov decision processes
- regret bounds
- markov decision problems
- control policies
- function approximation
- reinforcement learning algorithms
- utility function
- state space
- optimal policy
- temporal difference
- dynamic programming
- linear regression
- online learning
- policy iteration
- partially observable
- lower bound
- finite state
- action selection
- infinite horizon
- action space
- learning algorithm
- learning process
- average reward
- bregman divergences
- control strategies
- control strategy
- information theoretic
- optimality criterion