Fast SVM classifier for large-scale classification problems

Huajun Wang, Genghui Li, Zhenkun Wang

Abstract: Highlights•The new SVM model. We construct a new SVM model, which is named as truncated squared hinge loss SVM model.•The new and efficient ADMM algorithm. We propose a new and effective alternating direction method of multipliers with working set for solving the truncated squared hinge loss SVM model.•High numerical performance. We compare our algorithm with eight other efficient solvers. Numerical experiments show that our algorithm shares excellent performance.