报告题目：Entropy-regularized Reinforcement Learning for Zero-Sum Stochastic Differential Games and the Application in Insurance
时 间：2025年12月26日（星期五）11:00
地 点：科研楼18号楼1102
主 办：数学与统计学院
参加对象：感兴趣的老师和学生
报告摘要：To overcome challenge form parameter misspecification in traditional stochastic differential games, this paper introduces a distributional control mechanism that characterizes optimal strategies as probability distributions dependent on both the current state and system parameters, rather than deterministic functions. This mechanism evolves into a reinforcement learning framework for Zero-Sum Stochastic Differential Games (RL-ZSSDGs) with jump diffusion processes. By applying the dynamic programming principle, the Hamilton-Jacobi-Bellman-Isaacs (HJBI) equations associated with the RL-ZSSDGs are derived. From these, expressions for the equilibrium strategies of the players are specified by the derivatives of the value function. To illustrate the framework, we applied it to linear quadratic problems and reinsurance-investment problems, and analyzed the effect of the learning rate on optimal policies and value functions.
报告人简介：李丹萍，华东师范大学统计学院教授，博士生导师，入选上海市“晨光计划”“东方英才计划青年项目（原上海市青年拔尖人才）”。主要研究方向为保险精算、金融数学、金融工程，在Mathematical Finance、Mathematics of Operations Research、Journal of Economic Dynamics and Control、Insurance: Mathematics and Economics、SIAM Journal on Financial Mathematics等国内外权威期刊发表学术论文40余篇，出版专著1部，撰写的专报获中央及省部级领导批示采纳。主持国家自然科学基金项目2项，荣获第八届高等学校科学研究优秀成果奖（人文社会科学）二等奖、天津市优秀博士学位论文等。担任中国优选法统筹法与经济数学研究会量化金融与保险分会副秘书长。