Sub2Vec: Feature learning for subgraphs

Abstract

Network embeddings have become very popular in learning effective feature representations of networks. Motivated by the recent successes of embeddings in natural language processing, researchers have tried to find network embeddings in order to exploit machine learning algorithms for mining tasks like node classification and edge prediction. However, most of the work focuses on distributed representations of nodes that are inherently ill-suited to tasks such as community detection which are intuitively dependent on subgraphs. Here, we formulate subgraph embedding problem based on two intuitive properties of subgraphs and propose Sub2Vec, an unsupervised algorithm to learn feature representations of arbitrary subgraphs. We also highlight the usability of Sub2Vec by leveraging it for network mining tasks, like community detection and graph classification. We show that Sub2Vec gets significant gains over state-of-the-art methods. In particular, Sub2Vec offers an approach to generate a richer vocabulary of meaningful features of subgraphs for representation and reasoning.