SIGN: Scalable Inception Graph Neural Networks

TL;DR

SIGN introduces a scalable graph neural network architecture that eliminates sampling, enabling fast training and inference on large graphs while maintaining competitive performance and achieving state-of-the-art results on massive datasets.

Contribution

The paper proposes a novel scalable GNN architecture using precomputable graph filters, avoiding sampling and enabling efficient training on large graphs.

Findings

Competitive performance on various benchmarks.

State-of-the-art results on ogbn-papers100M.

Significantly faster training and inference times.

Abstract

Graph representation learning has recently been applied to a broad spectrum of problems ranging from computer graphics and chemistry to high energy physics and social media. The popularity of graph neural networks has sparked interest, both in academia and in industry, in developing methods that scale to very large graphs such as Facebook or Twitter social networks. In most of these approaches, the computational cost is alleviated by a sampling strategy retaining a subset of node neighbors or subgraphs at training time. In this paper we propose a new, efficient and scalable graph deep learning architecture which sidesteps the need for graph sampling by using graph convolutional filters of different size that are amenable to efficient precomputation, allowing extremely fast training and inference. Our architecture allows using different local graph operators (e.g. motif-induced adjacency matrices or Personalized Page Rank diffusion matrix) to best suit the task at hand. We conduct extensive experimental evaluation on various open benchmarks and show that our approach is competitive with other state-of-the-art architectures, while requiring a fraction of the training and inference time. Moreover, we obtain state-of-the-art results on ogbn-papers100M, the largest public graph dataset, with over 110 million nodes and 1.5 billion edges.