This website requires JavaScript.

Bellman Meets Hawkes: Model-Based Reinforcement Learning via Temporal Point Processes

Chao QuXiaoyu TanSiqiao XueXiaoming ShiJames ZhangHongyuan Mei
摘要
We consider a sequential decision making problem where the agent faces theenvironment characterized by the stochastic discrete events and seeks anoptimal intervention policy such that its long-term reward is maximized. Thisproblem exists ubiquitously in social media, finance and health informatics butis rarely investigated by the conventional research in reinforcement learning.To this end, we present a novel framework of the model-based reinforcementlearning where the agent's actions and observations are asynchronous stochasticdiscrete events occurring in continuous-time. We model the dynamics of theenvironment by Hawkes process with external intervention control term anddevelop an algorithm to embed such process in the Bellman equation which guidesthe direction of the value gradient. We demonstrate the superiority of ourmethod in both synthetic simulator and real-world problem.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答