Customizing Graph Neural Networks using Path Reweighting

TL;DR

This paper introduces CustomGNN, a novel graph neural network that learns to reweight paths based on their semantics for specific tasks, improving accuracy and robustness over traditional GNNs.

Contribution

The paper proposes CustomGNN, which automatically learns task-specific path semantics to enhance GNN performance and address issues like over-smoothing and overfitting.

Findings

Achieves state-of-the-art accuracy on multiple datasets

Effectively filters task-irrelevant paths

Reduces over-smoothing and improves robustness

Abstract

Graph Neural Networks (GNNs) have been extensively used for mining graph-structured data with impressive performance. However, because these traditional GNNs do not distinguish among various downstream tasks, embeddings embedded by them are not always effective. Intuitively, paths in a graph imply different semantics for different downstream tasks. Inspired by this, we design a novel GNN solution, namely Customized Graph Neural Network with Path Reweighting (CustomGNN for short). Specifically, the proposed CustomGNN can automatically learn the high-level semantics for specific downstream tasks to highlight semantically relevant paths as well to filter out task-irrelevant noises in a graph. Furthermore, we empirically analyze the semantics learned by CustomGNN and demonstrate its ability to avoid the three inherent problems in traditional GNNs, i.e., over-smoothing, poor robustness, and overfitting. In experiments with the node classification task, CustomGNN achieves state-of-the-art accuracies on three standard graph datasets and four large graph datasets. The source code of the proposed CustomGNN is available at \url{https://github.com/cjpcool/CustomGNN}.