Provably Efficient Offline Reinforcement Learning for Partially Observable Markov Decision Processes.
Hongyi GuoQi CaiYufeng ZhangZhuoran YangZhaoran WangPublished in: ICML (2022)
Keyphrases
- reinforcement learning
- partially observable markov decision processes
- continuous state
- markov decision processes
- partially observable domains
- state space
- optimal policy
- policy search
- finite state
- partially observable environments
- partial observability
- function approximation
- belief state
- policy gradient
- hidden state
- planning under uncertainty
- temporal difference
- partially observable
- reinforcement learning algorithms
- machine learning
- transfer learning
- markov chain
- search space
- decision making