Multi-view subspace clustering via simultaneously learning the representation tensor and affinity matrix

•A novel multi-view subspace clustering method (GLTA) is proposed.•GLTA adopts the tensor nuclear norm to explore high-order correlation among multiple features.•GLTA exploited the manifold regularization to preserve the view-specific geometrical structures.•GLTA can automatically assign a optimal weight for each view.•Extensive experiments on seven real-world datasets are given for validation.