New primal SVM solver with linear computational cost for big data classifications

Abstract

Support Vector Machines (SVM) is among the most popular classification techniques in ma-chine learning, hence designing fast primal SVM algorithms for large-scale datasets is a hot topic in recent years. This paper presents a new L2-norm regularized primal SVM solver using Aug-mented Lagrange Multipliers, with linear com-putational cost for Lp-norm loss functions. The most computationally intensive steps (that de-termine the algorithmic complexity) of the pro-posed algorithm is purely and simply matrix-by-vector multiplication, which can be easily paral-lelized on a multi-core server for parallel com-puting. We implement and integrate our algo-rithm into the interfaces and framework of the well-known LibLinear software toolbox. Experi-ments show that our algorithm is with stable per-formance and on average faster than the state-of-the-art solvers such as SVMperf, Pegasos and the LibLinear that integrates the TRON, PCD and DCD algorithms. 1.