报告人：蔡剑锋（香港科技大学）

邀请人：刘海霞

报告时间：2021年7月6日（星期二）9:00-10:30

报告地点：腾讯会议：631 455 666

报告题目：Landscape analysis of non-convexoptimizations in phase retrieval

报告摘要：Non-convex optimization is a ubiquitous tool in scientific and engineering research. For many important problems,simple non-convex optimization algorithms often provide good solutions efficiently and effectively, despite possible local minima. One way to explain the success of these algorithms is through the global landscape analysis. In this talk, we present some results along with this direction for phase retrieval. The main results are, for several of non-convex optimizations in phase retrieval, a local minimum is also global and all other critical points have a negative directional curvature. The results not only will explain why simple non-convex algorithms usually find a global minimizer for phase retrieval, but also will be useful for developing new efficient algorithms with a theoretical guarantee by applying algorithms that are guaranteed to find a local minimum.

报告人简介：蔡剑锋，香港科技大学数学系教授。主要研究兴趣是数据科学和成像技术中的算法设计与分析。在包括JAMS，ACHA，SIAM系列期刊，IEEE Trans.系列期刊，JMLR，CVPR，ICCV等发表论文60多篇。