2012/03/13 by Anastasios Kyrillidis, Volkan Cevher, Kyrillidis, Anastasios +1
Computer Science · Engineering · #Blind Source Separation Techniques #Distributed Sensor Networks and Detection Algorithms #FOS: Computer and information sciences #Information Theory (cs.IT) #Sparse and Compressive Sensing Techniques
paper · pdf · doi:10.48550/arxiv.1203.2936
openalex publication_date 2012/03/13 · openalex created_date 2025/10/24 · openalex updated_date 2026/07/28
The least absolute shrinkage and selection operator (LASSO) for linear\nregression exploits the geometric interplay of the \ℓ2-data error\nobjective and the \ℓ1-norm constraint to arbitrarily select sparse models.\nGuiding this uninformed selection process with sparsity models has been\nprecisely the center of attention over the last decade in order to improve\nlearning performance. To this end, we alter the selection process of LASSO to\nexplicitly leverage combinatorial sparsity models (CSMs) via the combinatorial\nselection and least absolute shrinkage (CLASH) operator. We provide concrete\nguidelines how to leverage combinatorial constraints within CLASH, and\ncharacterize CLASH's guarantees as a function of the set restricted isometry\nconstants of the sensing matrix. Finally, our experimental results show that\nCLASH can outperform both LASSO and model-based compressive sensing in sparse\nestimation.\n