Do We Need Anisotropic Graph Neural Networks?

Shyam A. Tailor, Felix Opolka, Pietro Lio, Nicholas Donald Lane

Keywords: graph neural networks, efficiency, latency reduction, memory reduction, architecture design, benchmarking, hardware-aware

Abstract: Common wisdom in the graph neural network (GNN) community dictates that anisotropic models---in which messages sent between nodes are a function of both the source and target node---are required to achieve state-of-the-art performance. Benchmarks to date have demonstrated that these models perform better than comparable isotropic models---where messages are a function of the source node only. In this work we provide empirical evidence challenging this narrative: we propose an isotropic GNN, which we call Efficient Graph Convolution (EGC), that consistently outperforms comparable anisotropic models, including the popular GAT or PNA architectures by using spatially-varying adaptive filters. In addition to raising important questions for the GNN community, our work has significant real-world implications for efficiency. EGC achieves higher model accuracy, with lower memory consumption and latency, along with characteristics suited to accelerator implementation, while being a drop-in replacement for existing architectures. As an isotropic model, it requires memory proportional to the number of vertices in the graph ($\mathcal{O}(V)$); in contrast, anisotropic models require memory proportional to the number of edges ($\mathcal{O}(E)$). We demonstrate that EGC outperforms existing approaches across 6 large and diverse benchmark datasets, and conclude by discussing questions that our work raise for the community going forward. Code and pretrained models for our experiments are provided at https://github.com/shyam196/egc.

One-sentence Summary: We find that using a simple adaptive filtering approach for GNNs improves performance over SOTA, while reducing model memory consumption and latency.