Fast Learning Algorithm For Support Vector Machine Based On Shell Vector

Support vector machine (SVM) method based on statistics learning theory has been widely researched and used for its good generalized ability, and has lots of important results. But SVM method still has some shortcomings, such as the training speed is very slow for large-scale data, etc.In this paper, support vector machine method is analyzed in detail, and some fast learning algorithms for support vector machines based on hull vetor are proposed. The main research in this paper can be shown as follows:1) A new geometric fast learning algorithm for support vector machine (SVM) is proposed. A set of hull vectors which are most likely to become the support vectors are extracted from training samples with the help of geometric information in these samples, the obtained hull vector set is considered as the new training sample set, which greatly reduces the time consumed in solving sequential quadratic optimization problems in SVM training, speeding up the training process. Simulations showed the algorithm is much faster than the normal SVM method and the classification accuracy is the same.2) On the basis of hull vector SVM, a new geometric fast incremental learning algorithm for support vector machine is proposed. A set of hull vectors which are most likely to become the support vector are extracted from training samples with the help of geometric information in these samples. In the incremental learning process, the obtained hull vector set and a new sample set are conjoined as the updated training sample set, which greatly reduces the time consumed in solving sequential quadratic optimization problems in incremental SVM training, speeding up the training process. In addition, compared with the existing incremental SVM learning algorithms in which only support vectors are used to represent original sample set, the proposed algorithm improves the classification precision. Simulations showed the effectiveness of the algorithm.3) The algorithms are applied in the automobile license plate recognition area.