ReFactor GNNs: Revisiting Factorisation-based Models from a Message-Passing Perspective

TL;DR

This paper introduces ReFactor GNNs, a novel architecture that unifies factorisation-based models and graph neural networks through message-passing, achieving strong inductive performance with fewer parameters.

Contribution

It reformulates factorisation models as GNNs via message-passing, enabling inductive learning and reducing parameter count.

Findings

ReFactor GNNs achieve state-of-the-art inductive performance.

ReFactor GNNs match transductive performance of FMs.

The model uses significantly fewer parameters.

Abstract

Factorisation-based Models (FMs), such as DistMult, have enjoyed enduring success for Knowledge Graph Completion (KGC) tasks, often outperforming Graph Neural Networks (GNNs). However, unlike GNNs, FMs struggle to incorporate node features and generalise to unseen nodes in inductive settings. Our work bridges the gap between FMs and GNNs by proposing ReFactor GNNs. This new architecture draws upon both modelling paradigms, which previously were largely thought of as disjoint. Concretely, using a message-passing formalism, we show how FMs can be cast as GNNs by reformulating the gradient descent procedure as message-passing operations, which forms the basis of our ReFactor GNNs. Across a multitude of well-established KGC benchmarks, our ReFactor GNNs achieve comparable transductive performance to FMs, and state-of-the-art inductive performance while using an order of magnitude fewer parameters.