Semi-supervised Learning Based On Anchor Graph Laplacian Regularization

Recently,research has be shown that,accurate recognition and classification of data sample points obtained through various ways is beneficial to data research work.In the actual research work,only a very small number of marked samples are often obtained from the data sample points,so it is very important to the feature information of the few labeled sample points to predict the labels.Based on this problem,two kinds of semi-supervised learning methods based on anchor graph structure are proposed as follows in this paper:(1)This paper aims at constructing coupled graph to discover the intrinsic sample structures under the Semi-Supervised Learning(SSL).Specifically,the graph Laplacian matrices over anchors and samples are respectively constructed by the weight matrix.The anchor graph gives the coarse data structure and reduces the influences of the noise of training samples and outliers.And the sample graph gives the detailed description for the fine structures of samples.Experiments on several publicly datasets show that the proposed approach achieves the superiorclassification performances,while the computational costs are comparable to stateof-the-art methods.(2)In the prediction of unlabeled sample points,by updating the unlabeled sample corresponding to the tag with the highest confidence after prediction into the labeled sample,the scale of the labeled sample is expanded.Therefore,based on the previous work,an improved joint Selflearning is proposed.By mining the effective information of anchor point structure and sample point structure on tag prediction,the accuracy of the final tag prediction matrix is improved.And the experimental results on the general data set show that,the prediction performance of this method is superior to that of other algorithms.