ContextGNN goes to Elliot: Towards Benchmarking Relational Deep Learning for Static Link Prediction (aka Personalized Item Recommendation)

TL;DR

By framing personalized item recommendation as static link prediction on a user-item graph and leveraging a k-hop subgraph $\mathcal{N}_k(u)$, the paper applies ContextGNN within Relational Deep Learning (RDL). The model computes a pair-wise score $y^{(pair)}(u,i) = \langle h_u^{(k)}, h_i^{(k)} \rangle$ for local items and a two-tower score $y^{(tower)}(u,i) = \langle h_u^{(k)}, q_i \rangle$ for others, fused by an adaptive mechanism. The authors integrate ContextGNN into the Elliot benchmarking platform with RelBench, reproduce results on Gowalla, Yelp 2018, and Amazon Book against six baselines, and provide a practical workflow for large-scale experiments. Overall, the results corroborate that ContextGNN often outperforms traditional MP-based baselines but may lag behind UltraGCN and GFCF in some settings, and the work delivers a reusable benchmarking pipeline to accelerate relational deep learning research for personalized item recommendation.

Abstract

Relational deep learning (RDL) settles among the most exciting advances in machine learning for relational databases, leveraging the representational power of message passing graph neural networks (GNNs) to derive useful knowledge and run predicting tasks on tables connected through primary-to-foreign key links. The RDL paradigm has been successfully applied to recommendation lately, through its most recent representative deep learning architecture namely, ContextGNN. While acknowledging ContextGNN's improved performance on real-world recommendation datasets and tasks, preliminary tests for the more traditional static link prediction task (aka personalized item recommendation) on the popular Amazon Book dataset have demonstrated how ContextGNN has still room for improvement compared to other state-of-the-art GNN-based recommender systems. To this end, with this paper, we integrate ContextGNN within Elliot, a popular framework for reproducibility and benchmarking analyses, counting around 50 state-of-the-art recommendation models from the literature to date. On such basis, we run preliminary experiments on three standard recommendation datasets and against six state-of-the-art GNN-based recommender systems, confirming similar trends to those observed by the authors in their original paper. The code is publicly available on GitHub: https://github.com/danielemalitesta/Rel-DeepLearning-RecSys.