时 间：2025年12月03日（周三）15：00 – 16：00
地 点：线上，腾讯会议号：147 242 434；密码：845396
报告人：张志民 重庆大学教授
主持人：李丹萍 华东师范大学教授
摘 要：
This study integrates variable annuities with a Guaranteed Lifetime Withdrawal Benefit (GLWB) into a life-cycle consumption and portfolio selection model to address the longevity risk, health risks, uncertainty in medical expenditures, and financial market risks faced by individuals before and after retirement. We construct a discrete-time life-cycle model that incorporates multiple sources of risk, including labor income, health status, medical expenses, and bequest motives, and use this model to introduce a multi stage decision-making problem involving GLWB annuities. To solve this high dimensional stochastic optimization problem, we model it as a finite-horizon Markov Decision Process (MDP) and propose a generalized Proximal Policy Optimization (PPO) algorithm suitable for finite time horizons. We derive a policy improvement lower bound under finite-horizon MDPs and design an efficient gradient estimation method to enhance sample efficiency.
报告人简介：
张志民，重庆大学教授、博士生导师，重庆市学术技术带头人，香港大学和墨尔本大学访问学者。目前担任中国工业与应用数学学会理事，中国双法研究会量化金融与保险分会常务理事、重庆市统计学会理事等。主要研究兴趣为保险精算、金融数学、统计学习、机器学习等。在本领域的权威杂志IME、AB、SAJ、AOR、中国科学（数学）等发表高水平论文80余篇。主持4项国家自科基金，多项省部级和横向课题。