Semi-Supervised Classification Of Self-Learning Association Matrix

In recent years,graph-based semi-supervised learning has attracted widespread attention from most researchers due to its good classification effect.It included two steps of constructing the correlation matrix and predicting labels of unlabeled samples,the association between paired samples can be effectively use which can effectively used to predict the label of unlabeled samples.Based on this idea,this paper proposes two semi-supervised classification methods of self-learning correlation matrix.The specific research contents and contributions of this paper are summarized as follows:?1?A semi-supervised classification method that based on ₁ norm simultaneous learning correlation matrix and Laplacian regularized least squares is proposed,the specific idea is that the sparse self-representation problem of samples and the Laplace regularization least-squares classifier are effectively fused together,a Self-taught Laplacian Regularized Least Square?ST-LapRLS?model was established.Simultaneous optimization and mutual improvement of the sample correlation matrix and Laplacian regularization least squares classifier were achieved.The validity of our method is fully verified on real datasets.?2?Based on the previous work,a semi-supervised classification method based on kernel norm self-learning correlation matrix is proposed.This method is more suitable for the semi-supervised Manifold Assumption.The proposed method performs well on four face datasets.