Abstract
Learning network embedding for large-scale networks have been attracting increasing attention due to their importance in supporting numerous network analytic and data mining tasks such as node classification, clustering and visualization. In this paper, we present a novel framework for learning large-scale network embedding incorporating network topology and community structural information. Most existing network embedding methods tend to embed network topology and ignore the partially labeled community structure information that exist in real-world networks and thus are unable to efficiently learn and capture the community structure of real-world networks. Unlike existing works, our framework integrates the network topology and community structure into the learning process. We propose a deep autoencoder model to generate low-dimensional feature representations efficiently through learning network reconstruction and community classification tasks. The experimental results on several real-world networks show that our framework outperforms the state-of-the-art methods.