Fast Computation of LOO-CV Function for SVM Classification

Support vector machine classification(SVC) is known to provide a more complete description of the linear and nonlinear relationships between input vectors and classifiers. Model selection of SVC is concerned about the choice of appropriate hyper-parameters which affect the performance of SVC. In this paper, we propose a fast computation method of a leave-one-out cross-validation(LOO-CV) function used in the model selection of SVC. Numerical results are then presented which illustrate the efficiency of the proposed computation method of LOO-CV function.