Generalized elastic net Lp-norm nonparallel support vector machine. (February 2020)

• Record Type: Journal Article
• Authors: Li, Chun-Na; Ren, Pei-Wei; Shao, Yuan-Hai; Ye, Ya-Fen; Guo, Yan-Ru
• Is Part Of: Engineering applications of artificial intelligence. Volume 88(2020)
• Publication Date: 2020-02
• Subjects: Support vector machine -- Nonparallel support vector machine -- Lp-norm -- Generalized elastic net -- Regularization; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2019.103397
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Generalized eigenvalue proximal support vector machine (GEPSVM) is the first nonparallel support vector machine. Compared to standard support vector machine (SVM), GEPSVM coped with the "Xor" problem well. In this paper, by defining a generalized elastic net regularization which is the combination of the L2-norm and Lq-norm, we propose a generalized elastic net Lp-norm nonparallel proximal support vector machine (GLpNPSVM), where p, q > 0 . GLpNPSVM measures the distance of a sample to each hyperplane by the Lp-norm, and hence can achieve desired performance by choosing appropriate p . In addition, the generalized elastic net regularization makes GLpNPSVM own good generalization ability. GLpNPSVM is a generalized formulation, and GEPSVM and some of its improvements are special cases of GLpNPSVM. A simple but effective iterative technique is introduced to solve GLpNPSVM, and we prove its convergence for certain p, q > 0 . Experimental results on different types of contaminated data sets show the effectiveness of GLpNPSVM.