subgraph2vec: Learning Distributed Representations of Rooted Sub-graphs from Large Graphs

TL;DR

subgraph2vec introduces an unsupervised method for learning vector representations of rooted subgraphs, enabling improved performance in graph classification, clustering, and real-world applications like code clone and malware detection.

Contribution

It presents a novel unsupervised approach to embed rooted subgraphs into vector space, enhancing graph analysis tasks with superior accuracy over existing methods.

Findings

Achieves significant accuracy improvements over existing graph kernels.

Effective in supervised and unsupervised graph learning tasks.

Outperforms state-of-the-art kernels in code clone and malware detection.

Abstract

In this paper, we present subgraph2vec, a novel approach for learning latent representations of rooted subgraphs from large graphs inspired by recent advancements in Deep Learning and Graph Kernels. These latent representations encode semantic substructure dependencies in a continuous vector space, which is easily exploited by statistical models for tasks such as graph classification, clustering, link prediction and community detection. subgraph2vec leverages on local information obtained from neighbourhoods of nodes to learn their latent representations in an unsupervised fashion. We demonstrate that subgraph vectors learnt by our approach could be used in conjunction with classifiers such as CNNs, SVMs and relational data clustering algorithms to achieve significantly superior accuracies. Also, we show that the subgraph vectors could be used for building a deep learning variant of Weisfeiler-Lehman graph kernel. Our experiments on several benchmark and large-scale real-world datasets reveal that subgraph2vec achieves significant improvements in accuracies over existing graph kernels on both supervised and unsupervised learning tasks. Specifically, on two realworld program analysis tasks, namely, code clone and malware detection, subgraph2vec outperforms state-of-the-art kernels by more than 17% and 4%, respectively.