报告人:Zhuo Jin金卓(澳大利亚Macquarie大学)
报告题目:A Reinforcement Learning Approach for Portfolio Selection with GMDBs
报告摘要:This paper addresses portfolio optimization for retired investors managing risk-free assets, risky assets, and variable annuities with GMDBs under mortality and surrender risks. Using the Lee-Carter model, it analyzes Australian demographic data, predicts mortality risk, and simulates surrender risk for fair GMDB pricing. A deep reinforcement learning (DRL) algorithm optimizes high-dimensional asset allocation, leveraging neural networks to adapt dynamically to market changes. The algorithm's global convergence is proven, ensuring robustness. Numerical experiments validate its effectiveness in managing complex portfolios, highlighting its stability and advantages in integrating mortality, surrender, and financial risks for retirement planning.
报告时间:2025年1月2日(星期四)9:30-11:00
报告地点:科技楼706会议室
邀请人:吴付科
报告人简介:金卓教授,澳大利亚麦考瑞大学精算中心教授,2005年和2007毕业于华中科技大学数学系应用数学专业,分别获理学学士和硕士学位,2011年毕业于美国韦恩州立大学数学系数学专业,获哲学博士学位。2011年至2022年在澳大利亚墨尔本大学经济系精算中心工作,2022年至今当前在澳大利亚麦考瑞大学精算中心工作。研究方向为随机最优控制,随机系统的数值方法,精算学,数理金融。在国际期刊发表70余篇论文,期刊包括Insurance Mathematics and Economics, European Journal of Operational Research, SIAM Journal on Control and Optimization, Automatica, ASTIN: Bulletin, Scandinavian Actuarial Journal。