Mitigating Overfitting in Graph Neural Networks via Feature and Hyperplane Perturbation

TL;DR

This paper introduces a novel data augmentation method for graph neural networks that flips initial features and hyperplanes to reduce overfitting caused by feature sparsity, significantly improving classification accuracy.

Contribution

It presents the first approach to mitigate overfitting in GNNs due to feature sparsity by augmenting data through feature and hyperplane flipping.

Findings

Increases node classification accuracy by up to 46.5%.

Enhances robustness of GNNs on real-world datasets.

Addresses overfitting caused by feature sparsity in GNNs.

Abstract

Graph neural networks (GNNs) are commonly used in semi-supervised settings. Previous research has primarily focused on finding appropriate graph filters (e.g. aggregation methods) to perform well on both homophilic and heterophilic graphs. While these methods are effective, they can still suffer from the sparsity of node features, where the initial data contain few non-zero elements. This can lead to overfitting in certain dimensions in the first projection matrix, as training samples may not cover the entire range of graph filters (hyperplanes). To address this, we propose a novel data augmentation strategy. Specifically, by flipping both the initial features and hyperplane, we create additional space for training, which leads to more precise updates of the learnable parameters and improved robustness for unseen features during inference. To the best of our knowledge, this is the first attempt to mitigate the overfitting caused by the initial features. Extensive experiments on real-world datasets show that our proposed technique increases node classification accuracy by up to 46.5% relatively.