Deep Fusion Clustering Network

Deep clustering is a fundamental yet challenging task for data analysis. Recently we witness strong tendency of combining autoencoder and graph neural networks to exploit structure information performance enhancement. However, observe that existing literature 1) lacks dynamic fusion mechanism selectively integrate refine the node attributes consensus representation learning; 2) fails extract from both sides robust target distribution (i.e., “groundtruth” soft labels) generation. To tackle above issues, propose Fusion Clustering Network (DFCN). Specifically, in our network, an interdependency learning-based Structure Attribute Information (SAIF) module proposed explicitly merge representations learned by learning. Also, reliable generation measure triplet self-supervision strategy, which facilitate cross-modality exploitation, are designed network training. Extensive experiments on six benchmark datasets have demonstrated DFCN consistently outperforms state-of-the-art deep methods.