Subspace Clustering With Two-Dimensional Graph Pca And Discrete Label Learning

Subspace clustering has been widely used in many areas today. Traditional subspace clustering methods convert data into vectors, which may damage structure information of data and lead to performance decline. In this paper, we not only find a projection matrix and project graph data into projection directions for preserving the structure information, but also perform discrete transforms on intermediate continuous labels after getting adjacency matrix to get the discrete clustering labels. Experiments conducted on significant data sets have confirmed the priority of our method comparing with some classic traditional methods.