A Fast Incremental Learning Algorithm for SVM Based on K Nearest Neighbors

A fast incremental learning algorithm for SVM based on K nearest neighbors (KNN-ISVM) is proposed. The algorithm extracts border vector set by applying the idea of K nearest neighbors and trains SVM by substituting the border vectors set for training set. The method can reduce training samples and speeds up training process. By adjusting value of K, useful training samples can be reserved farthest in the border vector set and the ability of SVM is improved. The experiment results demonstrate the effectivity of KNN-ISVM.