报告题目一：Convergence of an Inexact ADMM for Separable Convex Optimization

报告人：   张洪超 副教授

报告时间：8月21日（周二）14:30-16:30

报告地点：3044am永利集团欢迎您学术报告厅  

报告摘要：In this talk we will discuss some convergence properties of an inexact ADMM method, where the subproblems are solved inexactly to an adaptive accuracy. We will talk about the global convergence as well as its convergence complexities under different situations. In general, we will show that this inexact ADMM method will maintain the same convergence complexity as the corresponding exact ADMM method, where all the subproblems need to be solved exactly. Preliminary numerical result shows the benefits of solving the subproblems inexactly by the new proposed method.

报告题目二：Inexact proximal stochastic gradient method for convex composite optimization

报告时间：8月21日（周二）16:30-18:30

报告地点：3044am永利集团欢迎您学术报告厅  

报告摘要: We will talk about an inexact proximal stochastic gradient (IPSG) method for convex composite optimization, whose objective function is a summation of an average of a large number of smooth convex functions and a convex, but possibly nonsmooth, function. Variance reduction techniques are incorporated in the method to reduce the stochastic gradient variance. The main feature of this IPSG algorithm is to allow solving the proximal subproblems inexactly while still keeping the global convergence with desirable complexity bounds. We will talk in more detail about the convergence properties of this IPSG. Some numerical results will be also discussed.

报告人简介：张洪超，博士，路易斯安那州立大学数学系和计算中心副教授。研究兴趣：非线性优化理论和应用，稀疏矩阵，医学成像中的反问题，无导数优化等，近年来主持多项美国国家自然科学基金(NSF)项目。在Mathematical Programming、SIAM Journal on Optimization、SIAM Journal on Numerical Analysis、SIAM Journal on Scientific Computing、SIAM Journal on Imaging、SIAM Journal on Control and Optimization等优化计算领域顶尖期刊发表论文20余篇。现为Computational Optimization and Applications （计算优化与应用，Springer，二区）、Optimization Letters等期刊编委。

欢迎广大师生参加！

3044am永利集团欢迎您

2018年8月16日