Preconditioning strategies for a nested primal-dual method with application to image deblurring
2024年11月20日,星期日,10:30-11:30
讲座题目:Preconditioning strategies for a nested primal-dual method with application to image deblurring
讲座时间 Datetime: 2024年11月20日,星期日,10:30-11:30
地点 Venue: 中大珠海校区海琴二号A457
报告人 Speaker: Donatelli Marco 教授
单位 Affiliation: DiSAT, University of Insubria (Italy)
主持人 Host: Davide Bianchi 副教授
报告摘要 Abstract:
Abstract: Proximal-gradient methods are widely used in imaging and can be accelerated by adopting variable metrics and/or extrapolation steps. However, one crucial issue is the inexact computation of the proximal operator, often implemented through a nested primal-dual solver. This represents the main computational bottleneck, especially when higher accuracy is required.
In this talk, we present preconditioning strategies for a nested primal-dual method aimed at efficiently solving regularized convex optimization problems. For our preconditioner inspired to the Iterated Tikhonov method, we prove the convergence of the iterates sequence towards a solution of the problem. Numerical results confirm that our preconditioned nested primal-dual method effectively accelerates iterations towards the solution of the problem, with reduced computational cost per iteration, especially in image deblurring problems with total variation regularization.
报告人简介:
马尔科·多纳泰利(Marco Donatelli),1977年出生,2002年在佛罗伦萨大学获得计算机科学“Laurea”学位,2006年在米兰大学获得应用数学博士学位。自2005年以来,他一直在科莫的因苏布里亚大学科学与高科技系工作,自2023年起担任数值分析全职教授。他是因苏布里亚大学计算机科学与计算数学博士项目的副协调员。
他的研究兴趣包括多重网格方法、逆问题的数值方法、图像和信号恢复的正则化方法、谱分析和预处理。他与约100位合作者合作,发表了100多篇论文,主要在国际期刊上发表。他还担任多个国际期刊的编辑委员会成员,包括NLAA、AMC和ETNA。此外,他还是2022年国家研究项目PRIN“成像科学中的逆问题 (IPIS)”的本地协调员。