报告题目：A generalized alternating direction implicit method for consensus optimization with application to distributed sparse logistic regression
报 告 人: 张文星 副教授 电子科技大学
报告时间：2021 年10 月28 日 上午 10:00 - 10:40
报告地点：腾讯会议 ID：959 559 128 会议密码：9999
校内联系人：李欣欣 xinxinli@jlu.edu.cn
报告摘要：A large family of paradigmatic models arising in the area of image/signal processing, machine learning and statistics regression can be boiled down to consensus optimization problems. This paper is devoted to handling generic consensus optimization by reformulating it as a monotone inclusion problem. We extend the algorithmic framework of the Hermitian and skew-Hermitian splitting (HSS) method for linear systems of equations to the monotone plus skew-symmetric circumstance. Under some mild conditions, the proposed algorithm converges globally and is favourable for tackling consensus optimization on distributed computing architecture. Numerical experiments on sparse logistic regression are implemented in two distributed fashions. Compared to some state-of-the-art methods such as the alternating direction method of multipliers and the stochastic variance reduced gradient, the novel method exhibits competitive and appealing performances, especially when its relaxation factor approaches to zero.
报告人简介：张文星，电子科技大学副教授，2012年博士毕业于南京大学数学系。2014-2015年在法国图卢兹大学从事博士后研究。曾访问香港浸会大学、香港大学等高校。主要研究兴趣为变分不等式、最优化理论与算法、及其在信息科学中的应用。主持国家自然科学基金面上项目一项。在Math Comput, Inverse Problems, J Sci Comput, SIAM J Imaging Sci, IEEE Trans Medical Imag, IEEE Trans Image Process, J Math Image Vis等杂志发表论文20余篇。
