PyTorch Geometric (PyG) Graph Learning

About

This skill provides specialized expertise in PyTorch Geometric (PyG), the premier library for geometric deep learning. It enables Claude to assist with the entire GNN lifecycle, from defining complex graph data structures and implementing message-passing paradigms to training state-of-the-art architectures like GCN, GAT, and GraphSAGE. Use this skill to navigate irregular data structures, optimize mini-batching for large-scale graphs, and apply graph machine learning to domains such as bioinformatics, social network analysis, and recommendation systems.

Key Features

• Implementation of standard GNN layers including GCN, GAT, and GraphSAGE
• Integration with benchmark datasets like Planetoid, TUDataset, and QM9
• Guidance on custom message-passing and neighborhood aggregation schemes
• Efficient mini-batch processing and large-scale graph dataset handling
• 3,719 GitHub stars
• Support for heterogeneous graphs and multi-type node/edge relations

Use Cases

• Predicting molecular properties and drug interactions in chemical research
• Building high-performance citation network classifiers and recommendation engines
• Performing social network analysis for community detection and influence modeling