Pessimistic Bootstrapping for Uncertainty-Driven Offline Reinforcement Learning.
Chenjia BaiLingxiao WangZhuoran YangZhi-Hong DengAnimesh GargPeng LiuZhaoran WangPublished in: ICLR (2022)
Keyphrases
- reinforcement learning
- expected utility
- possibility theory
- partial observability
- data driven
- function approximation
- uncertain data
- optimal policy
- sequential decision problems
- model free
- optimal control
- utility function
- reinforcement learning algorithms
- state space
- policy search
- neural network
- probability theory
- weakly supervised
- machine learning
- action selection
- multi agent
- learning process
- incomplete information
- artificial intelligence
- named entity recognition
- hidden markov models
- learning capabilities
- uncertain information
- temporal difference learning
- data sets
- multi agent reinforcement learning
- decision makers
- real time