Beyond Low-Pass Filters: Adaptive Feature Propagation on Graphs

TL;DR

This paper introduces an adaptive GNN model that leverages spectral filters and attention mechanisms to effectively handle both homophilic and heterophilic graphs, outperforming existing methods especially on heterophilic data.

Contribution

The paper proposes a novel flexible GNN model with spectral attention that generalizes across different graph types, overcoming the homophily assumption of traditional GNNs.

Findings

Outperforms state-of-the-art on heterophilic graphs

Performs comparably on homophilic graphs

Generalizes well across diverse graph structures

Abstract

Graph neural networks (GNNs) have been extensively studied for prediction tasks on graphs. As pointed out by recent studies, most GNNs assume local homophily, i.e., strong similarities in local neighborhoods. This assumption however limits the generalizability power of GNNs. To address this limitation, we propose a flexible GNN model, which is capable of handling any graphs without being restricted by their underlying homophily. At its core, this model adopts a node attention mechanism based on multiple learnable spectral filters; therefore, the aggregation scheme is learned adaptively for each graph in the spectral domain. We evaluated the proposed model on node classification tasks over eight benchmark datasets. The proposed model is shown to generalize well to both homophilic and heterophilic graphs. Further, it outperforms all state-of-the-art baselines on heterophilic graphs and performs comparably with them on homophilic graphs.