Obtaining accurate estimated action values in categorical distributional reinforcement learning.
Yingnan ZhaoPeng LiuChenjia BaiWei ZhaoXianglong TangPublished in: Knowl. Based Syst. (2020)
Keyphrases
- obtaining accurate
- reinforcement learning
- action selection
- attribute values
- numerical values
- function approximation
- numerical data
- partially observable domains
- dynamic programming
- standard deviation
- action space
- human actions
- neural network
- optimal policy
- learning process
- markov decision processes
- user defined
- action recognition
- categorical attributes
- agent learns
- multi agent
- machine learning