Statistics and Samples in Distributional Reinforcement Learning
作者：互联网
摘要：郑重声明：原文参见标题，如有侵权，请联系作者，将会撤销发布！
arXiv:1902.08102v1 [stat.ML] 21 Feb 2019
Abstract
1. Introduction
2. Background
2.1. Bellman equations
2.2. Categorical and quantile distributional RL
CDRL.
QDRL.
3. The role of statistics in distributional RL
3.1. Expectiles
3.2. Imputation strategies
3.3. Expectile distributional reinforcement learning
3.4. Stochastic approximation
4. Analysing distributional RL
4.1. Bellman closedness
4.2. Approximate Bellman closedness
4.3. Mean consistency
5. Experimental results
5.1. Tabular policy evaluation
EDRL.
QDRL.
5.2. Tabular control
5.3. Expectile regression DQN
6. Conclusion
标签：Statistics,distributional,Distributional,QDRL,Bellman,Expectile,closedness,Reinf 来源： https://www.cnblogs.com/lucifer1997/p/13303865.html