Unsupervised learning from attributed networks

Unlike the plain network where only the topological structure is available, nodes of attributed networks possess rich attributed information. Thereby, the attributed networks ubiquitous in the real world have attracted much attention in recent years. In our paper, to simultaneously address both attributed network embedding and clustering, we propose a new model. It exploits both content and structure/topological information, capitalizing on their simultaneous use. The proposed model relies on the approximation of the relaxed continuous embedding solution by the true discrete clustering. Thus, we show that incorporating an embedding representation provides simpler and more interpretable solutions. Moreover, we show how our proposed approach is related to some other clustering and data embedding methods. To clearly establish various connections, we propose to distinguish two variants. Experimental results demonstrate that the proposed algorithm performs better, in terms of clustering and embedding, than the state-of-the-art algorithms, including deep learning methods dedicated to similar tasks for attributed network datasets.