Learning Graph Representations With Global Structural Information

As the era of Internet Plus is coming, more and more data is produced per day. So it is the duty of data mining to extract useful information from the big data. In reality, data is always organized by graph structure, which is an important issues of graph mining. Among many different algorithms, recently proposed graph representations based methods has been arrived state-of-the-art performance, and also has more focus in research area.In this paper, we present GraRep, a novel model for learning vertex representations of weighted graphs. This model learns low dimensional vectors to represent vertices appearing in a graph and, unlike existing work, integrates global structural information of the graph into the learning process. We also formally analyze the connections between our work and several previous research efforts, including the Deep Walk model of Perozzi et al. as well as the skip-gram model with negative sampling of Mikolov et al.We conduct experiments on a language network, a social network as well as a citation net-work and show that our learned global representations can be effectively used as features in tasks such as clustering, classification and visualization. Empirical results demonstrate that our representation significantly outperforms other state-of-the-art methods in such tasks.However, there were still several shortcomings. As for the huge calculation on matrix multi-plication and singular value decomposition, it would be very slow when encountering large-scale data sets. In future works, we will investigate the approximation algorithm for matrix multiplication and online algorithm design, as well as alternatives of singular value decom-position for data dimensions reduction.