Towards Safe Policy Improvement for Non-Stationary MDPs.
Yash ChandakScott M. JordanGeorgios TheocharousMartha WhitePhilip S. ThomasPublished in: NeurIPS (2020)
Keyphrases
- non stationary
- finite horizon
- optimal policy
- markov decision processes
- markov decision problems
- markov decision process
- policy search
- reinforcement learning
- state and action spaces
- state space
- policy iteration
- average cost
- average reward
- infinite horizon
- stock price
- partially observable
- temporal evolution
- adaptive algorithms
- autoregressive
- reward function
- action space
- linear programming
- empirical mode decomposition
- policy evaluation
- reinforcement learning problems
- factored mdps
- optimal control
- white noise
- finite state
- dynamic programming
- computer vision