主题：A Transfer Learning Framework for Efficient Portfolio Selection
主讲人：林红梅 上海对外经贸大学
主持人：王国长 暨南大学
时间：2026年9月30日（周三）下午14:30-15:30
地点：暨南大学石牌校区经济学院（中惠楼）503室
摘要
Portfolio selection plays a critical role in financial decision-making by helping investors optimize returns, meet financial objectives, and manage risk. However, practical applications often face challenges such as limited data and high noise levels, particularly in emerging markets. This paper investigates the potential of transfer learning to address these issues in portfolio selection. We propose a general framework that incorporates information from auxiliary sources through a proposed transferable source detection algorithm. We further provide a rigorous theoretical analysis, establishing $l_1$-norm estimation error bounds and consistency guarantees for the proposed methods. Empirical results on both simulated and real-world datasets demonstrate that our transfer learning approach substantially improves portfolio performance, yielding higher Sharpe ratios and more efficient asset allocation, especially when more source data are available or a logarithmic utility function is employed.
主讲人简介
林红梅，上海对外经贸大学统计与数据科学学院教授，博士生导师；入选上海市曙光计划、东方英才青年项目。主要研究方向为非参半参回归分析、函数型数据分析以及分布式统计方法等，在JASA,JCGS,Sinica等发表论文30余篇，主持多项国家自然科学基金、上海市自然科学基金。主要学术兼职包括：担任中国现场统计研究会常务理事，中国现场统计研究会教育统计与管理分会副理事长。
校对 |欧阳萍
责编 | 彭毅
初审 | 姜云卢
终审发布 | 何凌云
（来源：暨南大学经济学院微信公众号）
