Near-Minimax-Optimal Risk-Sensitive Reinforcement Learning with CVaR.
Kaiwen WangNathan KallusWen SunPublished in: ICML (2023)
Keyphrases
- risk sensitive
- optimal control
- risk neutral
- reinforcement learning
- model free
- markov decision processes
- risk averse
- optimality criterion
- control policies
- dynamic programming
- utility function
- evaluation function
- markov decision problems
- infinite horizon
- worst case
- control strategy
- supervised learning
- temporal difference
- optimal solution
- reinforcement learning algorithms
- average cost
- average reward
- finite state
- finite horizon
- reward function
- radial basis function
- decision problems
- decision makers
- state space
- multi agent