Uncertainty-Based Offline Reinforcement Learning with Diversified Q-Ensemble.
Gaon AnSeungyong MoonJang-Hyun KimHyun Oh SongPublished in: NeurIPS (2021)
Keyphrases
- reinforcement learning
- partial observability
- learning algorithm
- temporal difference
- function approximation
- ensemble learning
- ensemble methods
- feature selection
- neural network
- random forest
- incomplete information
- robotic control
- neural network ensemble
- temporal difference learning
- ensemble classifier
- state space
- training data
- uncertain data
- supervised learning
- training set
- partially observable
- reinforcement learning algorithms
- multi class
- real robot
- learning problems
- multi agent
- inherent uncertainty
- policy search
- ensemble members
- markov decision processes
- probability distribution
- robust optimization
- belief functions
- classifier ensemble
- optimal control
- base classifiers
- expected utility
- support vector machine
- model free
- decision theory