A unified algorithm framework for mean-variance optimization in discounted Markov decision processes.
Shuai MaXiaoteng MaLi XiaPublished in: Eur. J. Oper. Res. (2023)
Keyphrases
- markov decision processes
- dynamic programming
- model based reinforcement learning
- policy iteration
- state space
- average reward
- linear programming
- optimal policy
- finite state
- np hard
- probabilistic model
- computational complexity
- learning algorithm
- continuous state spaces
- finite horizon
- markov decision process
- total reward
- long run
- decision theoretic planning
- factored mdps
- state abstraction
- probabilistic planning
- reinforcement learning algorithms
- convergence rate
- search space
- search algorithm