A ReLU-based Hard-thresholding Algorithm for Non-negative Sparse Signal Recovery

报告人：温金明教授（暨南大学）
报告时间： 5月11日上午10:00-12:00
腾讯会议：441 954 964

报告摘要：
   In numerous applications, such as DNA microarrays, face recognition and spectral unmixing, we need to acquire a non-negative K-sparse signal x from an underdetermined linear model y= Ax+v. To recover such sparse signals, we propose a ReLU-based hard-thresholding algorithm (RHT) and then develop two sufficient conditions of stable recovery with RHT, which are respectively based on the restricted isometry property (RIP) and mutual coherence of the sensing matrix A. As far as we know, these two sufficient conditions are the best for hard-thresholding-type algorithms. Finally, we perform extensive numerical experiments to show that RHT has better overall recovery performance and more efficient than the non-negative least squares (NNLS) algorithm, some hard-thresholding-type algorithms including the iterative hard-thresholding (IHT) algorithm, hard-thresholding pursuit (HTP), Newton-step-based iterative hard-thresholding algorithm (NSIHT) and Newton-step-based hard-thresholding pursuit (NSHTP), and Non-Negative orthogonal matching pursuit (NNOMP), Fast NNOMP (FNNOMP) and Support-Shrinkage NNOMP (SNNOMP), which are variants of orthogonal matching pursuit (OMP) for recovering non-negative sparse signals.

报告人介绍：
   温金明，暨南大学教授、博导、国家高层次青年人才、广东省青年珠江学者，主持国家自然科学基金面上项目2项，省级项目4项；2015年6月博士毕业于加拿大麦吉尔大学数学与统计学院。从2015年3月到2018年9月，温教授先后在法国科学院里昂并行计算实验室、加拿大阿尔伯塔大学、多伦多大学从事博士后研究工作。温教授研究方向是整数信号和稀疏信号恢复的算法设计与理论分析，以第一作者/通讯作者在Applied and Computational Harmonic Analysis、IEEE Transactions on Information Theory、IEEE Transactions on Signal Processing等期刊和会议发表50余篇学术论文。