An iterative SVM approach to feature selection and classification in high-dimensional datasets

The support vector machine (SVM) is the state-of-the-art classification method, and the doubly regularized SVM (DrSVM) is its important extension based on the elastic net penalty. DrSVM has been successfully applied in handling variable selection and while retaining (or discarding) correlated variables. However, it is challenging to solve this model. In this paper we develop an iterative l2-SVM approach to implement DrSVM over high dimensional datasets via an alternating direction method of multipliers (ADMM). Our approach can significantly reduce the computation complexity and further accelerate the convergence speed. Moreover, the corresponding algorithms have global convergence property. Experimental results over the simulated and real-world gene datasets are encouraging.