Speaker: Prof. Zengjing Chen (陈增敬 教授)

Time: 15:00-15:50, 14 March 2025 (Friday)  (Beijing time)

Venue: T2-202, UIC

Language: Chinese

Abstract

Formulating algorithms to solve constrained high-dimensional Hamilton-Jacobi-Bellman (HJB) equations has long been challenging, largely due to the inherent complexities associated with the constrained condition and the “curse of dimensionality”. Artificial Intelligence (AI), by mimicking human cognition, has significantly advanced the resolution of open problems across various fields, as exemplified by the “cap set problem” with “Fun-Search”. This work pioneers a novel AI-based approach by transforming the challenge of solving general constrained high-dimensional HJB equations into a problem of optimizing strategies with multiple four-armed slot machines. This approach leverages an equivalence between certain stochastic control problems and the multi-armed slot machine framework, recasting finite-region control as a policy optimization challenge over an infinite strategy set. It represents a groundbreaking application of AI methodologies to a classical class of partial differential equations, with promising potential for broad applications in fields such as finance, engineering, and physics.

About the Speaker

陈增敬，山东大学教授，山东大学中泰证券金融研究院院长，山东国家应用数学中心执行委员会常务副主任。主要从事金融数学、倒向随机微分方程、非线性期望、计量经济学等领域的研究，先后在Econometrica、Journal of Economic Theory、Annals of Probability、Automatica 和Nature 子刊等期刊发表论文 80 余篇。在量子和非独立的框架下，给出了一类非线性正态分布分布密度的显示表达式。丰富和完善了彭实戈院士的非线性期望理论，并应用到金融领域，解决了资产定价领域中一些长期未解决的难题，在国内外产生了重要的影响。其中，与美国艺术与科学院士、著名经济学家 Epstein 合作发现了动态多先验资产定价理论与非线性g-期望之间的联系，得到了被称为 Chen-Epstein 的定价公式。该结果被诺贝尔经济奖获得者、国际数学家 (ICM) 报告人以及多位国际著名学者和专家引用或推广。曾先后获得第十四届孙冶方经济科学奖、国家自然科学二等奖和“五一”劳动奖章等诸多奖项。目前正在主持国家重点研发项目一项、山东省自然科学基金重大基础研究项目一项；曾主持国家杰出青年科学基金、国家自然科学基金重点项目各一项。