Abstract:Over the past two decades, support vector machine (SVM) has become a popular supervised machine learning model, and plenty of distinct algorithms are designed separately based on different KKT conditions of the SVM model for classification/regression with different losses, including the convex loss or nonconvex loss. In this paper, we propose an algorithm that can train different SVM models in a \emph{unified} scheme. First, we introduce a definition of the \emph{LS-DC} (\textbf{l}east \textbf{s}quares type of \textbf{d}ifference of \textbf{c}onvex) loss and show that the most commonly used losses in the SVM community are LS-DC loss or can be approximated by LS-DC loss. Based on DCA (difference of convex algorithm), we then propose a unified algorithm, called \emph{UniSVM}, which can solve the SVM model with any convex or nonconvex LS-DC loss, in which only a vector is computed, especially by the specifically chosen loss. Particularly, for training robust SVM models with nonconvex losses, UniSVM has a dominant advantage over all existing algorithms because it has a closed-form solution per iteration, while the existing algorithms always need to solve an L1SVM/L2SVM per iteration. Furthermore, by the low-rank approximation of the kernel matrix, UniSVM can solve the large-scale nonlinear problems with efficiency. To verify the efficacy and feasibility of the proposed algorithm, we perform many experiments on some small artificial problems and some large benchmark tasks with/without outliers for classification and regression for comparison with state-of-the-art algorithms. The experimental results demonstrate that UniSVM can achieve comparable performance in less training time. The foremost advantage of UniSVM is that its core code in Matlab is less than 10 lines; hence, it can be easily grasped by users or researchers.

A nonlinear kernel SVM classifier via L 0 / 1 soft-margin loss with classification performance

Using Svm To Model And Control Nonlinear Dynamical Systems

Generalized Convexity-Based Inexact Projection Method for Multiple Kernel Learning

Nonlinear kernel-free quadratic hyper-surface support vector machine with 0-1 loss function

A Novel Svm Modeling Approach For Highly Imbalanced And Overlapping Classification

Linear Programming-Based Sparse Kernel Regression with L1-Norm Minimization for Nonlinear System Modeling

A Novel Loss Function-based Support Vector Machine for Binary Classification

A Semismooth-Newton's-Method-Based Linearization and Approximation Approach for Kernel Support Vector Machines

Unified Locally Linear Classifiers with Diversity-Promoting Anchor Points

Non-smooth optimization algorithm to solve the LINEX soft support vector machine

An ADMM Solver for the MKL-$L_{0/1}$-SVM

A Quantum Extension of SVM-perf for Training Nonlinear SVMs in Almost Linear Time

A Non-Biased Form of Least Squares Support Vector Classifier and Its Fast Online Learning

Svm with Quadratic Polynomial Kernel Function Based Nonlinear Model One-Step-Ahead Predictive Control

Unified SVM Algorithm Based on LS-DC Loss

A Feature Selection Method for Nonparallel Plane Support Vector Machine Classification

Online Kernel Learning with a Near Optimal Sparsity Bound

Design of a Multiple Kernel Learning Algorithm for LS-SVM by Convex Programming

A Robust Optimization Model for Nonlinear Support Vector Machine

An Efficient Algorithm for a Class of Large-Scale Support Vector Machines Exploiting Hidden Sparsity.

Multikernel semiparametric linear programming support vector regression