Augmented Markov decision process Q-Learning(AMDP-Q) was proposed,which was inspired by the thought of combination of AMDP, Monte Carlo-partially observable Markov decision process(MC-POMDP) and Q-learning.Firstly a lower-dimensional sufficient statistic was taken to represent the belief state space.In the common situation,a good choice was the tuple of the maximum likelihood state and the entropy of the belief.The new space composed of this tuple was referred to as the augmented state space.Secondly,a set ...
| 
推荐 
点击下列相关研究机构和相关文献作者，可以直接查到这些机构和作者被《中国知识资源总库》收录的其它文献，使您全面了解该机构和该作者的研究动态和历史。
