Sparse support vector machines with L0 approximation for ultra-high dimensional omics data
作者：
Highlights：
• L0 SVM for simultaneous feature selection and classification.
• L0 SVM outperformed L1 SVM in that it identifies true features more accurately with less false positive rate.
• Implemented with dual optimization and is highly computationally efficient with ultra-high dimensional data (n ≪ p).
• Identifying biological important features with metagenomic and expression data.
• More sparse estimators with less bias.
摘要
•L0 SVM for simultaneous feature selection and classification.•L0 SVM outperformed L1 SVM in that it identifies true features more accurately with less false positive rate.•Implemented with dual optimization and is highly computationally efficient with ultra-high dimensional data (n ≪ p).•Identifying biological important features with metagenomic and expression data.•More sparse estimators with less bias.
论文关键词：SVM,L0 approximation,Ultra-high dimensional data,Feature selection,Classification,Metagenomics sequencing
论文评审过程：Received 24 May 2018, Revised 31 March 2019, Accepted 27 April 2019, Available online 30 April 2019, Version of Record 8 May 2019.
论文官网地址：https://doi.org/10.1016/j.artmed.2019.04.004
