报告人:段雪峰 教授
报告题目:A New Tensor Multi-rank Approximation with Total Variation Regularization for Tensor Completion
报告时间:2026年5月3日(周日)下午4:00
报告地点:云龙校区6号楼304报告厅
主办单位:数学与统计学院、数学研究院、科学技术研究院
报告人简介:
段雪峰,博士,二级教授,博士生导师,八桂学者,广西杰出青年基金获得者。主要从事机器学习、数值代数研究,近年来主持国家自然科学基金5项,广西自然科学基金3项,以第一作者或通讯作者在本领域权威期刊《Proceedings of the American Mathematical Society》、《Advances in Computational Mathematics》、《BIT Numerical Mathematics》、《IMA Journal of Numerical Analysis》和《Journal of Scientific Computing》等上发表SCI论文80余篇,其中一二区25篇,ESI高被引论文1篇。以第一完成人获得广西自然科学二等奖和广西高等教育教学成果二等奖。曾访问美国威廉玛丽学院数学系、意大利国际数学物理中心和中科院计算数学所等学术机构,多次被邀请在国际会议上做大会报告。担任中国高等教育学会教育数学专业委员会副秘书长,广西运筹学会副理事长等。
报告摘要:
In this talk, we present a novel tensor completion model which combines the Laplace function and an anisotropic total variation regularization. The Laplace function is utilized to approximate the tensor multi-rank, and the total variation regularization is added to improve the local piecewise smoothness and preserve the edges of the restored tensor data. An efficient alternating direction method of multipliers is proposed to tackle the tensor completion model, and its convergence theorem is also derived. Extensive experimental results on color images, videos, multispectral images and magnetic resonance imaging data show the efficiency and effectiveness of the proposed method.