Attentive multi-view deep subspace clustering net

In this paper, we propose a novel Attentive Multi-View Deep Subspace Nets (AMVDSN), which deeply explores underlying consistent and view-specific information from multiple views fuse them by considering each view’s dynamic contribution obtained attention mechanism. Unlike most multi-view subspace learning methods that they directly reconstruct data points on raw or only consider consistency complementarity when representation in deep shallow space, our proposed method seeks to find joint latent explicitly considers both consensus among views, then performs clustering learned representation. Besides, different contribute differently learning, therefore introduce mechanism derive weight for view, much better than previous fusion the field of clustering. The algorithm is intuitive can be easily optimized just using Stochastic Gradient Descent (SGD) because neural network framework, also provides strong non-linear characterization capability compared with traditional approaches. experimental results seven real-world sets have demonstrated effectiveness against some state-of-the-art