Node Identifiers: Compact, Discrete Representations for Efficient Graph Learning

TL;DR

This paper introduces a framework for creating compact, discrete, and interpretable node identifiers using vector quantization, improving efficiency and interpretability in large-scale graph learning tasks.

Contribution

It proposes a novel end-to-end method for generating highly compact and discrete node representations that enhance efficiency and interpretability in graph neural network applications.

Findings

Achieves 6-15 dimensional node representations.

Significantly reduces memory and computation costs.

Maintains competitive accuracy across multiple graph tasks.

Abstract

We present a novel end-to-end framework that generates highly compact (typically 6-15 dimensions), discrete (int4 type), and interpretable node representations, termed node identifiers (node IDs), to tackle inference challenges on large-scale graphs. By employing vector quantization, we compress continuous node embeddings from multiple layers of a Graph Neural Network (GNN) into discrete codes, applicable under both self-supervised and supervised learning paradigms. These node IDs capture high-level abstractions of graph data and offer interpretability that traditional GNN embeddings lack. Extensive experiments on 34 datasets, encompassing node classification, graph classification, link prediction, and attributed graph clustering tasks, demonstrate that the generated node IDs significantly enhance speed and memory efficiency while achieving competitive performance compared to current state-of-the-art methods.