A sequential algorithm for sparse support vector classifiers

Support vector machines (SVMs), though accurate, are not preferred in applications requiring high classification speed or when deployed in systems of limited computational resources, due to the large number of support vectors involved in the model. To overcome this problem we have devised a primal SVM method with the following properties: (1) it solves for the SVM representation without the need to invoke the representer theorem, (2) forward and backward selections are combined to approach the final globally optimal solution, and (3) a criterion is introduced for identification of support vectors leading to a much reduced support vector set. In addition to introducing this method the paper analyzes the complexity of the algorithm and presents test results on three public benchmark problems and a human activity recognition application. These applications demonstrate the effectiveness and efficiency of the proposed algorithm.

►
► An algorithm for sparse SVM classifiers is proposed.
► The solution of general SVM is derived in the primal.
► Incremental and decremental selections of support vectors (SVs) are combined to approach the optimal solution.
► The size of the final SVM is controlled by a SV criterion and an upper-limit.
► It is more stable than LibSVM particularly for linear SVMs and produces SVMs of competitive performances.