Heterophily-informed Message Passing
This work addresses a key limitation in GNNs for graph-based tasks like classification and generative modeling, though it appears incremental as it builds on existing heterophily-aware methods.
The paper tackles the oversmoothing problem in graph neural networks (GNNs) by introducing a heterophily-informed message passing scheme that regulates message aggregation to preserve both low and high-frequency information, resulting in performance enhancements across datasets and architectures, including notable improvements in molecular generation benchmarks.
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.