This repository is an "awesome list" curating resources for self-supervised learning (SSL) on graphs, targeting researchers and practitioners in graph representation learning. It provides a comprehensive overview of SSL techniques, categorizing them into contrastive, generative, and predictive approaches, and offers a structured way to explore recent advancements and implementations.

How It Works

The project categorizes graph SSL methods into three main paradigms: contrastive learning (contrasting augmented views of graph data), generative learning (reconstruction or prediction of masked graph components), and predictive learning (using self-generated labels from graph properties or context). It also outlines common training strategies like pre-training/fine-tuning, joint learning, and unsupervised representation learning.

Quick Start & Requirements

This repository is a curated list of papers and code, not a runnable library. It requires no installation. Users can explore the linked papers and code repositories for specific implementations.

Highlighted Details

• Extensive categorization of graph SSL methods, including detailed breakdowns of contrastive learning (e.g., same-scale vs. cross-scale contrasting) and generative/predictive approaches.
• Summaries of methodology, implementation details, common datasets, and available open-source codes for numerous graph SSL techniques.
• Includes a comprehensive table comparing various methods across graph properties, pretext tasks, data augmentation, objective functions, and training strategies.
• Provides links to papers and GitHub repositories for many of the listed methods.

Maintenance & Community

The repository is maintained by LirongWu and Haitao Lin, with contact information provided for feedback and contributions. It cites a TKDE paper, indicating academic backing.

Licensing & Compatibility

The repository itself is a curated list and does not have a specific license. Individual code repositories linked within will have their own licenses.

Limitations & Caveats

As a curated list, this repository does not provide a unified API or framework. Users must individually assess and integrate the code from linked repositories, which may have varying dependencies, licenses, and maintenance statuses. Some listed methods lack readily available code.