Graphite: Iterative Generative Modeling of Graphs
This work addresses the challenge of modeling discrete and combinatorial graph data for machine learning applications, offering a novel framework that could benefit researchers and practitioners in graph-based domains.
The authors tackled the problem of learning representations for nodes in large graphs by proposing Graphite, a deep latent variable generative model that uses graph neural networks and an iterative graph refinement strategy, achieving superior performance on density estimation, link prediction, and node classification tasks compared to competing approaches.
Graphs are a fundamental abstraction for modeling relational data. However, graphs are discrete and combinatorial in nature, and learning representations suitable for machine learning tasks poses statistical and computational challenges. In this work, we propose Graphite, an algorithmic framework for unsupervised learning of representations over nodes in large graphs using deep latent variable generative models. Our model parameterizes variational autoencoders (VAE) with graph neural networks, and uses a novel iterative graph refinement strategy inspired by low-rank approximations for decoding. On a wide variety of synthetic and benchmark datasets, Graphite outperforms competing approaches for the tasks of density estimation, link prediction, and node classification. Finally, we derive a theoretical connection between message passing in graph neural networks and mean-field variational inference.