基于统一测量和张量学习的多视图无监督特征选择OA

Multi-view data are increasingly common in high-dimensional industrial applications.Feature selection is important as it preserves the original meaning and interpretability of features.This is especially useful when labels are scarce,making multi-view unsupervised feature selection(MvUFS)highly practical.Existing methods often fall short in exploring inter-view relationships and consistency.To overcome this shortage,this paper proposed a new method called SMUMT.It integrated self-representation learning to improve sample representation.It also used joint learning to build a reliable similarity graph for gui-ding feature selection.Additionally,it introduced tensor learning to model high-order correlations across views.It conducted clustering experiments on seven public datasets.Results show that SMUMT outperforms six state-of-the-art methods in most cases.It performed particularly well on image datasets.These findings confirm that this method is effective for feature selection and improves clustering performance.