报告题目: Randomized Algorithms with Successive Over-Relaxation for Tensor Completion
报 告 人: 喻高航 教授
报告时间: 2025年6月30日(星期一)10:00-11:00
地 点:37号楼3A02
邀 请 人: 潘少华教授
欢迎广大师生前往!
数学学院
2025年6月26日
报告摘要:
Tensor completion has attracted significant interest due to its broad applications in areas such as image repair, video retrieval, and analysis of multidimensional datasets. Nonetheless, current approaches often grapple with limited recovery results and inefficiencies in computation, particularly when dealing with substantial missing data or large datasets. This talk will introduce some randomized approach for tensor completion, which combines randomized subspace power iteration with successive Over-Relaxation (SOR) iteration technique. Within the Tucker decomposition structure, the proposed method initially utilizes randomized subspace power iteration to achieve a low-rank tensor approximation. It subsequently applies an adaptive over-relaxation method that merges the current estimate with the low-rank approximation at each step, enhancing the recovery process and speeding up convergence. Tests conducted on various datasets illustrate that the proposed randomized method excels in performance for large-scale tensor completion problems.
报告人简介:
喻高航,浙江科技大学教授、博导,主要从事张量数据分析、大规模优化计算及其在机器学习、图像处理与医学影像中的应用研究。先后在SIAM Journal on Imaging Sciences, IEEE Transactions on Computational Social Systems,IEEE Signal Processing Letters,Expert Systems with Applications,Knowledge-Based Systems,Journal of Scientific Computing,Applied Mathematical Modelling,Inverse Problems, Computational Optimization and Applications, Journal of Optimization Theory and Applications, Optimization Methods and Software等国际期刊上发表50余篇SCI论文,先后主持5项国家自然科学基金、1项教育部新世纪优秀人才支持计划项目和1项浙江省自然科学基金重大项目,有多篇论文入选ESI高被引榜单。现任国际SCI学术期刊Intelligent Automation & Soft Computing 的期刊编委;国际学术期刊Statistics, Optimization and Information Computing执行编委(Coordinating Editor)。