A New SVM Active Learning Algorithm Based on KNN

Of all the active learning research, the study on the active learning algorithm for SVM is much less. In this paper, based on K-Nearest Neighbors (KNN), we propose a new SVM active learning algorithm. The algorithm first collects the potential informative samples to form a potential informative sample set, and then in this sample set, based on KNN it evaluates the sparseness for each sample. The sample that locates at a sparser region is taken as an informative one, and is selected for training. Experimental results show that the proposed algorithm can greatly improve the classification performance, and can efficiently accelerate the convergence of the classifier.