应002资讯网吴勃英、张达治老师邀请，上海交通大学数学科学学院张小群教授作学术报告，欢迎感兴趣的师生参加！

【报告题目】：Stochastic Primal Dual Fixed-Point Method for composite optimization and Applications in Data science

【报告时间】：2020年6月18日，下午14：00

【报告平台】：腾讯会议

点击链接直接加入会议：https://meeting.tencent.com/s/xBuXFpTvnAnH

【会议ID】：350 742 155

【报告摘要】：Many problems in machine learning and imaging sciences can be formulated a composite optimization problem. We consider a stochastic primal dual fixed-point method (SPDFP) and its variance reduction version (SVRG-PDFP) for solving the sum of two proper lower semi-continuous convex function and one of which is composite. The method is based on the primal dual fixed-point method (PDFP) that does not require subproblem solving. Under some mild condition, the convergence is established based on some standard assumptions. In particular, for SPDFP, the rate of the expected error of iterate is of the order $\mathcal{O}(k^{-\alpha})$ where $k$ is iteration number and $\alpha \in (0,1]$.  For SVRG-PDFP, we established the linear convergence for strongly convex problem and $\mathcal{O}(k^{1})$ for general convex function. Finally, numerical experiments on graphic Lasso, graphics logistic regressions and image reconstruction are provided to demonstrate the effectiveness of the proposed algorithm.

【报告人简介】：张小群,上海交通大学教授。武汉大学本科硕士，法国南布列塔尼大学应用数学博士。2007-2010年在美国加州大学洛杉矶分校任访问助理教授，2010加入上海交通大学自然科学研究院和数学科学学院。入选教育部新世纪优秀人才和青年拔尖人才计划。现任Inverse problems and Imaging杂志编委，中国工业与应用数学学会大数据与人工智能专业委员会委员。研究方向：图像科学，机器学习，数据科学与最优化。