Partial Policy Iteration for L1-Robust Markov Decision Processes.
Chin Pang HoMarek PetrikWolfram WiesemannPublished in: J. Mach. Learn. Res. (2021)
Keyphrases
- markov decision processes
- policy iteration
- finite state
- transition matrices
- optimal policy
- average reward
- dynamic programming
- model free
- reinforcement learning
- factored mdps
- state space
- sample path
- fixed point
- markov decision process
- least squares
- approximate dynamic programming
- partially observable
- policy evaluation
- markov decision problems
- action space
- decision processes
- planning under uncertainty
- policy iteration algorithm
- finite horizon
- infinite horizon
- average cost
- discounted reward
- reinforcement learning algorithms
- actor critic
- optimal control
- state and action spaces
- stochastic games
- temporal difference
- markov games
- cost function
- optical flow