Pairwise Learning for Neural Link Prediction

TL;DR

This paper introduces PLNLP, a flexible pairwise learning framework for neural link prediction that maximizes ranking metrics and achieves top performance on multiple benchmark datasets.

Contribution

The paper proposes a novel, adaptable pairwise learning framework for neural link prediction, integrating various neural architectures, scoring functions, and sampling strategies.

Findings

Achieves top 1 performance on ogbl-ddi and ogbl-collab datasets.

Achieves top 2 performance on ogbl-citation2 dataset.

Effective ranking loss improves link prediction accuracy.

Abstract

In this paper, we aim at providing an effective Pairwise Learning Neural Link Prediction (PLNLP) framework. The framework treats link prediction as a pairwise learning to rank problem and consists of four main components, i.e., neighborhood encoder, link predictor, negative sampler and objective function. The framework is flexible that any generic graph neural convolution or link prediction specific neural architecture could be employed as neighborhood encoder. For link predictor, we design different scoring functions, which could be selected based on different types of graphs. In negative sampler, we provide several sampling strategies, which are problem specific. As for objective function, we propose to use an effective ranking loss, which approximately maximizes the standard ranking metric AUC. We evaluate the proposed PLNLP framework on 4 link property prediction datasets of Open Graph Benchmark, including ogbl-ddi, ogbl-collab, ogbl-ppa and ogbl-ciation2. PLNLP achieves top 1 performance on ogbl-ddi and ogbl-collab, and top 2 performance on ogbl-ciation2 only with basic neural architecture. The performance demonstrates the effectiveness of PLNLP.