关键词: L1-范数距离, 鲁棒L1范数非平行近似支持向量机, 梯度上升, 线性方程, 分类问题

Abstract: Considering that Robust L1-norm non-parallel proximal support vector machine (L1-NPSVM) solves the resulted objective using an imperfect procedure of gradient ascending such that there is no guarantee of obtaining a more good solution,this paper proposes a novel iterative algorithm to solve the objective of L1-NPSVM. To be specific, the problem is transformed into a maximization problem with an equality constraint, such thati in each iterative, only two linear systems of linear equations needs to be solved. Theoretically, the convergence of this algorithm is proved. Experiments tried out on public UCI datasets show that the proposed algorithm obtains higher performance than L1-NPSVM and has comparable computing time.

Key words: L1-norm distance, Robust L1-norm non-parallel proximal support vector machine, Gradient ascending, Linear equations, Classification