Task-driven Heterophilic Graph Structure Learning

2025-12-29

AI Summary

To address the limited representational capacity of graph neural networks (GNNs) on heterogeneous graphs—caused by high label disparity and weak feature similarity among node neighborhoods—this paper proposes Frequency-guided Graph Structure Learning (FgGSL). FgGSL is the first framework to jointly integrate spectral-domain analysis with supervised graph topology inference, simultaneously modeling homophilic and heterophilic edges. It introduces a spectral encoder incorporating learnable symmetric feature masks and complementary high-pass/low-pass graph filters to enable task-aware structural optimization. Additionally, it defines a label-driven structural loss and provides theoretical guarantees on stability and robustness under graph perturbations. Evaluated on six mainstream heterogeneous graph benchmarks, FgGSL consistently outperforms state-of-the-art GNNs and graph rewiring methods, demonstrating the effectiveness and generalizability of coupling frequency-domain guidance with supervised structural learning.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

Graph neural networks (GNNs) often struggle to learn discriminative node representations for heterophilic graphs, where connected nodes tend to have dissimilar labels and feature similarity provides weak structural cues. We propose frequency-guided graph structure learning (FgGSL), an end-to-end graph inference framework that jointly learns homophilic and heterophilic graph structures along with a spectral encoder. FgGSL employs a learnable, symmetric, feature-driven masking function to infer said complementary graphs, which are processed using pre-designed low- and high-pass graph filter banks. A label-based structural loss explicitly promotes the recovery of homophilic and heterophilic edges, enabling task-driven graph structure learning. We derive stability bounds for the structural loss and establish robustness guarantees for the filter banks under graph perturbations. Experiments on six heterophilic benchmarks demonstrate that FgGSL consistently outperforms state-of-the-art GNNs and graph rewiring methods, highlighting the benefits of combining frequency information with supervised topology inference.

Problem

• Improves GNNs for heterophilic graphs with dissimilar connected nodes
• Learns complementary homophilic and heterophilic graph structures jointly
• Uses frequency-guided and task-driven inference to enhance node representation

Innovation

• Jointly learns homophilic and heterophilic graph structures with spectral encoder
• Uses learnable feature-driven masking and pre-designed filter banks
• Task-driven structural loss promotes edge recovery for heterophilic graphs