Riemannian Optimization for Low Rank Tensor Completion
数学专题报告
报告题目(Title):Riemannian Optimization for Low Rank Tensor Completion
报告人(Speaker):魏轲 研究员(复旦大学)
地点(Place):腾讯会议ID: 671-773-315
时间(Time):2023 年 02 月 23 日(周四) 10:00-11:00
邀请人(Inviter):蔡永强
报告摘要
This talk is about the low rank tensor completion problem, which is about reconstructing a tensor from partially revealed entries. Riemannian optimization algorithms based on the Tucker decomposition as well as the tensor train decommission will be discussed. On the one hand, exact recovery guarantee of the vanilla Riemannian gradient algorithm based on the Tucker decomposition will be established. On the other hand, we will discuss the quotient geometry of the low tensor train rank tensors under a preconditioned metric, together with the corresponding Riemannian optimization algorithms.
主讲人简介
魏轲, 复旦大学大数据学院青年研究员, 博士生导师. 2014年获得牛津大学博士学位, 之后在香港科技大学(2014-2015)和加州大学戴维斯分校(2015-2017)从事博士后研究. 主要研究兴趣为高维结构化数据处理算法与理论, 多智能体强化学习算法与理论, 数据科学的数理基础. 其研究成果已发表在国际重要的应用数学和工程期刊上,包括 SIAM系列、IEEE系列、Appl. Comput. Harmonic Anal.、Math. Program.、J Machine Learn Research、Inverse Problem等. 先后入选上海市扬帆计划、上海市特聘教授计划(东方学者)、中组部“青年拔尖人才计划”等,主持国家重点研发计划课题.