The Benefits of Being Distributional: Small-Loss Bounds for Reinforcement Learning.
Kaiwen WangKevin ZhouRunzhe WuNathan KallusWen SunPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- temporal difference learning
- loss bounds
- expert advice
- function approximation
- state space
- co occurrence
- small number
- worst case
- temporal difference
- probabilistic model
- fixed point
- linear regression
- game playing
- pairwise
- machine learning
- mutual information
- evaluation function
- active learning
- decision trees
- learning algorithm