报告人：徐姿教授

报告题目：Single-loop Optimization Algorithms for Nonconvex Minimax Optimization Problems

摘要：Much recent research effort has been directed to the development of efficient algorithms for solving nonconvex minimax problems with theoretical convergence guarantees due to the relevance of these problems to a few emergent applications, e.g., machine learning. In this talk, we first introduce some existing optimization algorithms for solving nonconvex minimax problems. Then, we propose a unified single-loop alternating gradient projection  (AGP) algorithm for solving nonconvex-(strongly) concave and (strongly) convex-nonconcave minimax problems. AGP employs simple gradient projection steps for updating the primal and dual variables alternatively at each iteration. To the best of our knowledge, the complexity results for solving the (strongly) convex-nonconcave minimax problems have never been obtained before in the literature. Moreover, we propose a zeroth-order alternating randomized gradient projection (ZO-AGP) algorithm for smooth nonconvex-concave minimax problems, and its iteration complexity is also analyzed. To the best of our knowledge, this is the first time that zeroth-order algorithms with iteration complexity guarantee are developed for solving both general smooth and block-wise nonsmooth nonconvex-concave minimax problems. Numerical results on data poisoning attack problem validate the efficiency of the proposed algorithms.

报告人简介：徐姿，上海大学理学院数学系，教授，博士生导师。中国运筹学会数学规划分会理事。主要研究方向是最优化理论与方法及在机器学习、无线通信等领域中的应用。研究成果在SIAM J. Optim.、IEEE JSAC等国际著名期刊上发表论文30余篇。主持多项国家自然科学基金项目和上海市自然基金项目。现任Springer 旗下优化期刊J. Global Optim.客座编委（Guest Editor）；担任国际期刊Numerical Algebra, Control & Optimization编委。曾赴美国明尼苏达大学、香港中文大学、香港大学、中国科学院、北京大学国际数学研究中心等机构学术访问和交流。2020年获得中国运筹学会科学技术奖青年科技奖。

报告时间：2021年11月3日14:00-16:00

报告方式：腾讯会议会议号：722 999 014