Develops and trains Graph Neural Networks (GNNs) for deep learning on irregular structures and relational data.
PyTorch Geometric (PyG) is a specialized skill that empowers Claude to architect, implement, and optimize Graph Neural Networks for complex data structures. It provides comprehensive guidance on node and graph classification, link prediction, and geometric deep learning using popular architectures like GCN, GAT, and GraphSAGE. Whether you are working on molecular discovery, social network analysis, or 3D point clouds, this skill streamlines the creation of data loaders, custom message-passing layers, and large-scale training workflows using the PyG framework.
主要功能
010 GitHub stars
02Implementation of popular GNN architectures like GCN, GAT, and GraphSAGE
03Custom message-passing layer construction using the MessagePassing base class
04Integration with benchmark datasets including Planetoid, QM9, and TUDataset
05Efficient mini-batch processing and data loading for large-scale graphs
06Support for heterogeneous graphs and multi-type node/edge relationships
使用场景
01Analyzing social networks for community detection and influence prediction
02Building recommendation systems based on user-item interaction graphs
03Predicting chemical properties and molecular structures for drug discovery
