A new twin SVM method with dictionary learning

Recently, dictionary learning has been widely studied, and lots of dictionary learning methods have been developed to solve the problem of classification. In this paper, we propose a new twin SVMs method with dictionary learning (TSVMDL) for classification. In the proposed method, we first incorporate the dictionary learning into twin SVMs to construct a unify model for prediction, in which we embed an analysis dictionary into learning that can obtain the coding coefficients and improve the representation ability of the dictionary. We further utilize the Lagrangian multiplier method to optimize the proposed TSVMDL objective model. We then obtain two nonparallel hyperplanes by solving two smaller sized quadratic programming problems (QPPs). Finally, extensive experiments have been conducted to evaluate the performance of the proposed TSVMDL method. The results have shown that our proposed method can obtain a better performance compared with state-of-the-art methods.

Automated summary

In this paper, a novel twin SVM method with dictionary learning (TSVMDL) was proposed to improve classification performance through better representation of coding coefficients. The proposed method incorporates analysis dictionary into twin SVMs and utilizes Lagrangian multiplier method for optimization to achieve better results compared to existing methods.