关键词: 支持向量机, 半监督学习, 预选取样本, 模糊C-均值聚类, 图像分类

Abstract: In order to solve the problems that the semi-supervised classification method based on Laplacian Support Vector Machines (LapSVM) requires that all unlabeled sample should be added to the training set to train a classifier, the algorithm demands high time and space and cannot effectively deal with large-scale image classification, Fuzzy C-Mean (FCM) pre-selected sample of LapSVM image classification method was proposed. The method used FCM algorithm for clustering the unlabeled samples. According to the clustering results, unlabeled samples of near optimal separating hyper-plane were selected to add to the training sample set, and these samples may be support vector carrying useful information for classification. The quantity was only a small part of the unlabeled samples, so the training sample set was reduced. The simulation results show that this method takes advantage of the inherent discrimination information of the unlabeled samples, effectively improves the accuracy of classifiers, and reduces the algorithm's time and space complexity.

Key words: Support Vector Machine (SVM), semi-supervised learning, pre-selected sample, Fuzzy C-means Clustering (FCM), image classification