Structural nonparallel support vector machine for pattern recognition. (December 2016)

• Record Type: Journal Article
• Authors: Chen, Dandan; Tian, Yingjie; Liu, Xiaohui
• Is Part Of: Pattern recognition. Volume 60(2016:Dec.)
• Page Start: 296
• Page End: 305
• Publication Date: 2016-12
• Subjects: Structural information -- Nonparallel support vector machine -- Alternating direction method of multipliers -- Pattern recognition; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2016.04.017
• Languages: English
• ISSNs: 0031-3203

Abstract

Abstract: It has been widely accepted that the underlying structural information in the training data within classes is significant for a good classifier in real-world problems. However, existing structural classifiers do not balance structural information׳s relationships both intra-class and inter-class. Combining the structural information with nonparallel support vector machine (NPSVM), we design a new structural nonparallel support vector machine (called SNPSVM). Each model of SNPSVM considers not only the compactness in both classes by the structural information but also the separability between classes, thus it can fully exploit prior knowledge to directly improve the algorithm׳s generalization capacity. Furthermore, we apply the improved alternating direction method of multipliers (ADMM) to SNPSVM. Both our model itself and the solving algorithm can guarantee that it can deal with large-scale classification problems with a huge number of instances as well as features. Experimental results show that SNPSVM is superior to the other current algorithms based on structural information of data in both computation time and classification accuracy. Abstract : Highlights: We design a new structural nonparallel support vector machine (SNPSVM). SNPSVM can fully exploit prior knowledge in the datasets. We apply the alternating direction method of multipliers (ADMM) for SNPSVM. We apply the block and parallel techniques in our algorithms.