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 <a class="link-external link-https" href="https://github.com/gim4855744/EFI-GNN" rel="external noopener nofollow">this https URL</a>

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that when existing graph neural networks (GNNs) process graph - structured data, they often only focus on the learning of high - order feature interactions and overlook the importance of low - order feature interactions. This design may lead to some valuable patterns being ignored, thereby affecting the prediction performance. To overcome this problem, the author proposes a new GNN method - Explicit Feature Interaction - aware Graph Neural Network (EFI - GNN), which can explicitly learn feature interactions of any order on the graph, thus making up for the deficiencies of existing GNNs. Specifically, the main contributions of the paper include: 1. **Proposing EFI - GNN**: This is a multi - layer linear network that explicitly models feature interactions of any order in the graph by performing feature cross - operation in each layer. Unlike traditional GNNs, EFI - GNN does not use activation functions but gradually increases the order of feature interactions through graph convolution and feature cross. 2. **Experimental verification**: The author conducted experiments using multiple datasets, and the results show that EFI - GNN is competitive with existing GNNs in prediction performance. Moreover, when EFI - GNN is jointly trained with other GNNs, the prediction performance is further improved. 3. **Interpretability**: Due to the linear structure of EFI - GNN, its prediction results are interpretable. The author shows the interpretability of the model by visualizing the heat maps of 1 - order and 2 - order feature interactions. 4. **Joint learning**: The author also explored the effect of jointly learning EFI - GNN with other GNNs. The experimental results show that combining explicit and implicit feature interactions can significantly improve the prediction performance. In conclusion, this paper aims to solve the limitations of existing GNNs in feature interaction learning by introducing EFI - GNN, thereby improving the processing effect of graph - structured data.

Explicit Feature Interaction-aware Graph Neural Networks

Explicit Feature Interaction-Aware Graph Neural Network

Discovering the Representation Bottleneck of Graph Neural Networks from Multi-order Interactions

GAIN: Graph Attention & Interaction Network for Inductive Semi-Supervised Learning over Large-scale Graphs

Discovering and Explaining the Representation Bottleneck of Graph Neural Networks from Multi-order Interactions

Bilinear Graph Neural Network with Neighbor Interactions

Dual Feature Interaction-Based Graph Convolutional Network

Utilizing Edge Features in Graph Neural Networks Via Variational Information Maximization

Beyond smoothness: A general optimization framework for graph neural networks with negative Laplacian regularization

Graph neural network with feature enhancement of isolated marginal groups

Interweaving Insights: High-Order Feature Interaction for Fine-Grained Visual Recognition

GoGNN: Graph of Graphs Neural Network for Predicting Structured Entity Interactions

Adaptive Kernel Graph Neural Network

BaGFN: Broad Attentive Graph Fusion Network for High-Order Feature Interactions

Enhancing Graph Neural Networks with Edge Features Through Sequential Representation

Cross-GCN: Enhancing Graph Convolutional Network with K-Order Feature Interactions

AKE-GNN: Effective Graph Learning with Adaptive Knowledge Exchange

Implicit Graph Neural Diffusion Networks: Convergence, Generalization, and Over-Smoothing

Efficient and Effective Implicit Dynamic Graph Neural Network

InteractionNN: A Neural Network for Learning Hidden Features in Sparse Prediction.

Multilevel Feature Interaction Learning for Session-Based Recommendation Via Graph Neural Networks