广州数学大讲坛第二期

第十六讲——深圳大学胡耀华教授学术报告

题目：Nonconvex Regularization Methods for Low Rank Matrix Recovery

时间：2026年3月6日（星期五）下午15:30-16:30

地点：理学实验楼312

报告人：胡耀华 教授

摘要：Low rank matrix recovery, aiming to find a low rank solution from the linear measurements, has a wide range of applications. In this talk, we will present the nonconvex regularization methods for the low rank matrix recovery problem in three aspects: theory, algorithm and application. In the theoretical aspect, by introducing a notion of spectral regularity condition, we will establish the global recovery bound for the nonconvex regularization problem. In the algorithmic aspect, we will apply the well-known proximal gradient method to solve the nonconvex regularization problem, and establish its linear convergence rate to the ground true low rank solution under a simple assumption. In the aspect of application, we will apply the nonconvex regularization method to solve genotype imputation problem for single-cell RNA-sequencing data.

报告人简介：

胡耀华，先后于浙江大学获得学士与硕士学位，香港理工大学获得博士学位。现任深圳大学数学科学学院特聘教授，副院长，博士生导师，香港理工大学兼职博导。主要从事连续优化理论、方法与应用研究，代表性成果发表在SIAM Journal on Optimization, Mathematical Programming, Mathematics of Operations Research, Inverse Problems, Journal of Machine Learning Research, Genome Biology, Bioinformatics等期刊，授权多项国家发明专利，开发多个生物信息学工具包与数据库。