WBT-BGRL: A Non-Contrastive Weighted Bipartite Link Prediction Model for Inductive Learning

TL;DR

This work tackles inductive link prediction in weighted bipartite graphs by proposing WBT-BGRL, a non-contrastive bootstrap framework that integrates edge weights into self-supervised learning with bipartite-specific encoders, dual predictors, and a momentum target network. It extends existing non-contrastive models to handle bipartite structure and weighing in both pretraining and link prediction, and provides a systematic ablation over four weighting configurations. Through experiments on Industry and E-commerce datasets with a strict inductive-temporal split, the study reveals dataset-dependent outcomes: edge weighting can improve generalization in skewed distributions but may cause overfitting, while a non-weighted pretraining regime can yield better generalization in highly skewed settings. The results highlight the value and limits of weighted, non-contrastive learning for inductive bipartite link prediction and offer practical design guidelines for when to apply weighting and how to structure encoders and predictors.

Abstract

Link prediction in bipartite graphs is crucial for applications like recommendation systems and failure detection, yet it is less studied than in monopartite graphs. Contrastive methods struggle with inefficient and biased negative sampling, while non-contrastive approaches rely solely on positive samples. Existing models perform well in transductive settings, but their effectiveness in inductive, weighted, and bipartite scenarios remains untested. To address this, we propose Weighted Bipartite Triplet-Bootstrapped Graph Latents (WBT-BGRL), a non-contrastive framework that enhances bootstrapped learning with a novel weighting mechanism in the triplet loss. Using a bipartite architecture with dual GCN encoders, WBT-BGRL is evaluated against adapted state-of-the-art models (T-BGRL, BGRL, GBT, CCA-SSG). Results on real-world datasets (Industry and E-commerce) show competitive performance, especially when weighting is applied during pretraining-highlighting the value of weighted, non-contrastive learning for inductive link prediction in bipartite graphs.