3.Multi-view semi-supervised feature selection with multi-order similarity and tensor learning

Multi-view data has attracted extensive attention because it can better characterize samples, and multi-view semi-supervised feature selection can not only effectively improve multi-view performance, but also maintain the original real structure of the data. To this end, many scholars have proposed various models to achieve this goal. However, most of the existing methods rely on the graph structure constructed from the original data and use the constructed graph as a guide for feature selection. This not only ignores multi-order domain knowledge, but also ignores the high-order relations between views. Therefore, this study effectively integrates multi-order domain information with graph learning, and performs tensor low-rank learning on the graph structure between multiple views. A multi-view semi-supervised feature selection method based on multi-order similarity and tensor learning is proposed, which not only integrates multi-order domain information, but also takes into account the relationship between views. Based on this, we propose an iterative method to solve the objective function and prove the superiority of our method on multiple basic datasets.

Multi-view semi-supervised feature selection with multi-order similarity and tensor learning, H.-Y. C, X.-J. X*, Y.-J. X, Neurocomputing, 2025, 657: 131573.