Feature Interaction-aware Graph Neural Networks
This addresses a bottleneck in graph learning for real-world applications with sparse data, though it appears incremental as an enhancement to existing GNN frameworks.
The paper tackles the problem of learning expressive node representations on graphs with high-dimensional and sparse features by proposing Feature Interaction-aware Graph Neural Networks (FI-GNNs), which highlight informative feature interactions in a personalized manner and demonstrate superior performance in experiments.
Inspired by the immense success of deep learning, graph neural networks (GNNs) are widely used to learn powerful node representations and have demonstrated promising performance on different graph learning tasks. However, most real-world graphs often come with high-dimensional and sparse node features, rendering the learned node representations from existing GNN architectures less expressive. In this paper, we propose \textit{Feature Interaction-aware Graph Neural Networks (FI-GNNs)}, a plug-and-play GNN framework for learning node representations encoded with informative feature interactions. Specifically, the proposed framework is able to highlight informative feature interactions in a personalized manner and further learn highly expressive node representations on feature-sparse graphs. Extensive experiments on various datasets demonstrate the superior capability of FI-GNNs for graph learning tasks.