报告题目:委托代理问题中的深度学习方法
报告人: 朱子木 助理教授 (香港科技大学(广州))
时间:2025年11月8日 上午9:00-10:30
地点:黑料网
创新港 涵英楼5-8001会议室
报告人简介:
朱子木, 现任香港科技大学(广州)金融科技方向助理教授。他于2021年获得南加州大学应用数学博士学位。加入香港科技大学(广州)之前,他曾于2021年至2024年担任加州大学圣巴巴拉分校客座助理教授。朱子木的研究兴趣包括随机控制与博弈、数学金融、委托代理问题和机器学习。他曾在《Mathematics of Operations Research》、《SIAM Journal on Financial Mathematics》和《Annals of Finance》等期刊上发表多篇论文。
摘要:
In this talk, I will discuss my joint work with M, Ludkovski and C.Xie.. Despite the importance of the Principal-Agent (PA) problem, there is a notable lack of relevant numerical algorithms. In the context of the PA problem, a significant subset of problems lacks explicit analytical solutions. This necessitates the development of a general numerical algorithm to address the PA problem. In this paper, we develop a novel Deep Galerkin Method (Deep PAAC) to solve the Hamilton-Jacobi-Bellman (HJB) partial differential equation arising in a general PA model. We investigate the role of the neural network architecture, training designs, loss functions, etc. on the convergence of the deep-learning solver and present 4 different case studies.