题目：Confined Orthogonal Matching Pursuit for Sparse Random Combinatorial Matrices

主讲人：温金明 教授

时间：2025年11月24日（周一） 9：30

地点：理学院楼310

主办单位：理学院

主讲人简介：

  温金明，吉林大学教授、博士生导师、国家青年人才、广东省青年珠江学者；现任中国数学会理事、广东省计算数学学会常务理事、广东省运筹学会常务理事、IEEE Trans. Audio Speech Lang. Process.、Alex. Eng. J.、《人工智能科学与工程》等期刊编辑。温教授的研究方向是整数信号和稀疏信号重构的算法设计与理论分析，近年来以第一作者/通讯作者和合作者在IEEE Trans. Inf. Theory、IEEE Trans. Signal Process.、IEEE/ACM Trans. Audio Speech Lang. Process.、ACM Trans. Asian Low-Resour. Lang. Inf. Process、SIAM J. Imaging Sci.、Inverse Probl.、Appl. Comput. Harmon. Anal.等期刊发表60余篇学术论文，以第一发明人授权中国发明专利15件。2020年至今连续6年入选全球前2%顶尖科学家。

摘要：

  This talk introduces a new Confined Orthogonal Matching Pursuit (Confined OMP) algorithm for sparse signal recovery over sparse random combinatorial matrices. The method constructs a confined set Γby exploiting the near-zero entries in the measurement vector and proves that, for a class of “confined signals” with i.i.d. nonzero components, the true support is contained inΓwith probability one. This property allows a substantial reduction of redundant columns in the measurement matrix, thereby significantly decreasing the identification complexity of the recovery algorithm. A lower bound on the exact recovery probability is further derived, showing that when the number of measurements satisfies m=2eKln(n−K), the recovery success probability is at least 1−1/(n−K). Simulation results demonstrate that the proposed method achieves orders-of-magnitude complexity reduction in low-sparsity regimes while maintaining competitive recovery performance and preserving robustness in noisy linear systems.