Provable Representation with Efficient Planning for Partially Observable Reinforcement Learning.
Hongming ZhangTongzheng RenChenjun XiaoDale SchuurmansBo DaiPublished in: CoRR (2023)
Keyphrases
- partially observable
- reinforcement learning
- state space
- markov decision processes
- decision problems
- partial observability
- dynamical systems
- partially observable domains
- markov decision problems
- belief state
- partially observable environments
- partial observations
- infinite horizon
- belief space
- fully observable
- hidden state
- planning domains
- reward function
- action models
- planning problems
- optimal policy
- action selection
- function approximation
- machine learning
- reinforcement learning algorithms
- dynamic programming
- initially unknown
- finite state
- orders of magnitude