Heterophily-informed Message Passing

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Haishan, Solin, Arno, Garg, Vikas
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911297344372736
author Wang, Haishan
Solin, Arno
Garg, Vikas
author_facet Wang, Haishan
Solin, Arno
Garg, Vikas
contents Graph neural networks (GNNs) are known to be vulnerable to oversmoothing due to their implicit homophily assumption. We mitigate this problem with a novel scheme that regulates the aggregation of messages, modulating the type and extent of message passing locally thereby preserving both the low and high-frequency components of information. Our approach relies solely on learnt embeddings, obviating the need for auxiliary labels, thus extending the benefits of heterophily-aware embeddings to broader applications, e.g., generative modelling. Our experiments, conducted across various data sets and GNN architectures, demonstrate performance enhancements and reveal heterophily patterns across standard classification benchmarks. Furthermore, application to molecular generation showcases notable performance improvements on chemoinformatics benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2504_19785
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Heterophily-informed Message Passing
Wang, Haishan
Solin, Arno
Garg, Vikas
Machine Learning
Graph neural networks (GNNs) are known to be vulnerable to oversmoothing due to their implicit homophily assumption. We mitigate this problem with a novel scheme that regulates the aggregation of messages, modulating the type and extent of message passing locally thereby preserving both the low and high-frequency components of information. Our approach relies solely on learnt embeddings, obviating the need for auxiliary labels, thus extending the benefits of heterophily-aware embeddings to broader applications, e.g., generative modelling. Our experiments, conducted across various data sets and GNN architectures, demonstrate performance enhancements and reveal heterophily patterns across standard classification benchmarks. Furthermore, application to molecular generation showcases notable performance improvements on chemoinformatics benchmarks.
title Heterophily-informed Message Passing
topic Machine Learning
url https://arxiv.org/abs/2504.19785