Elastic Net Collaborative Representation And Ensemble Learning Based Label Propagation Algorithm

Semi-supervised learning is an important research area in the field of machine learning The research of high dimensional semi-supervised learning algorithm is an important research topic in semi-supervised learning.With the rapid development of information technology,a large amount of data presents high dimensional characteristics in the fields of machine learning,pattern recognition,data mining and computer vision.Therefore,by implementing semi-supervised learning research on high dimensional data have important theoretical and practical value.In previous studies,based on the assumption that the data of the same class lie in the same subspace,the collaborative representation(CR)algorithm in semi-supervised learning has shown its well performance through label propagation by taking advantage of the fact that each labeled data can be well represented by the same class of unlabeled data.Unfortunately,the coefficients obtained by CR are not sparse,which means that most of the unlabeled data were used to reconstruct the labeled data,thus reducing the performance of the classifier.In order to solve this problem,an elastic net collaborative representation and ensemble learning based label propagation(ENCR-EL-LP)was proposed for high dimensional data.The proposed algorithm can be divided into the following steps:(1)utilizing Bootstrap sampling to select the feature subspace;(2)elastic net collaborative representation model is solved for each subspace;(3)using ensemble learning method to obtain the final representation coefficient;(4)utilizing the final representation coefficients to propagates the class information from labeled data to unlabeled dataIn order to verify the effectiveness of the proposed algorithm,experiments were carried out on the public data sets.The results show that the proposed algorithm ENCR-EL-LP achieved better classification accuracy than the semi-supervised learning based on PCA.