This repository serves as a comprehensive, curated collection of academic resources focusing on Graph Neural Networks (GNNs) that tackle the challenge of heterophily. It aims to provide researchers and practitioners with a centralized hub for papers and code addressing scenarios where connected nodes exhibit dissimilar features or labels, a departure from traditional GNN assumptions that favor homophily.

How It Works

The repository functions as a literature aggregator, meticulously listing research papers and their associated code. It highlights advancements in GNN architectures and methodologies designed to effectively model and leverage graph structures where the "opposites attract" phenomenon, or heterophily, is prevalent. This contrasts with the homophily assumption of many standard GNNs, which posit that connected nodes share similar attributes. The collection spans various approaches, from spectral methods to novel message-passing mechanisms.

Quick Start & Requirements

This repository is a curated list of research papers and code, not a runnable software package. Therefore, there are no direct installation or execution instructions provided. Users are expected to consult the linked papers for implementation details and specific dependencies.

Highlighted Details

• Features an extensive collection of academic papers and code from 2020 to 2024, covering a wide spectrum of GNN techniques specifically designed for heterophilic graphs.
• The resources delve into various facets of heterophily, including theoretical underpinnings, novel model designs, fairness considerations, robustness evaluations, and applications in node classification, link prediction, and anomaly detection.
• Includes contributions from leading research venues such as NeurIPS, ICLR, ICML, KDD, AAAI, and other prominent conferences and journals, offering a broad overview of the field's progress.

Maintenance & Community

The provided README does not contain information regarding repository maintenance, active community channels (like Discord or Slack), contributor statistics, or a public roadmap.

Licensing & Compatibility

No specific software license is mentioned in the README. Users are advised to refer to individual linked projects for their respective licensing details and compatibility notes.

Limitations & Caveats

As a collection of research pointers, this repository does not offer a ready-to-use GNN implementation. Users must independently access, evaluate, and potentially implement the referenced papers and code to address heterophily in their own projects, requiring significant effort to integrate and test.