Explicit Feature Interaction-aware Graph Neural Networks
Minkyu Kim, Hyun-Soo Choi, Jinho Kim
TL;DR
EFI-GNN addresses the limitation of traditional GNNs in capturing low-order feature interactions by introducing a multilayer linear architecture that explicitly learns arbitrary-order interactions on graphs via per-layer feature crossing without activation. The final model concatenates layer-wise representations and can be trained jointly with standard GNNs, yielding competitive accuracy and improved performance in joint settings, along with intrinsic interpretability through linear weights and feature-influence visualizations. Experiments across multiple datasets demonstrate the method's effectiveness and reveal that combining explicit (EFI-GNN) and implicit (GNN) interactions enhances predictive power. The work broadens the use of explicit feature interactions on graph-structured data and highlights opportunities for interpretable, high-order pattern learning in graphs.
Abstract
Graph neural networks (GNNs) are powerful tools for handling graph-structured data. However, their design often limits them to learning only higher-order feature interactions, leaving low-order feature interactions overlooked. To address this problem, we introduce a novel GNN method called explicit feature interaction-aware graph neural network (EFI-GNN). Unlike conventional GNNs, EFI-GNN is a multilayer linear network designed to model arbitrary-order feature interactions explicitly within graphs. To validate the efficacy of EFI-GNN, we conduct experiments using various datasets. The experimental results demonstrate that EFI-GNN has competitive performance with existing GNNs, and when a GNN is jointly trained with EFI-GNN, predictive performance sees an improvement. Furthermore, the predictions made by EFI-GNN are interpretable, owing to its linear construction. The source code of EFI-GNN is available at https://github.com/gim4855744/EFI-GNN