torch-geometric

Start with graph neural networks by installing PyG, loading a small graph into a Data object, and running a minimal two-layer GCN to observe graph convolution. This quick start approach demonstrates core concepts and training patterns in action.

Build a graph neural network for node classification by loading graph data into PyG Data structures, applying a two-layer GCN, and executing standard training patterns to classify nodes. The Skill provides practical examples covering this end-to-end workflow.

Yes, you can process heterogeneous graphs using PyG HeteroData structures. The framework provides built-in support for heterogeneous graph workflows, allowing you to model complex relationships and run learning tasks across different node and edge types.

Handle link prediction by leveraging PyG's graph data structures and transforms to construct training patterns. The framework supports link prediction workflows alongside node classification and graph classification, providing practical examples for evaluation.

Yes, PyG supports scalable training patterns for large graph datasets. It provides specialized data loaders and transforms that allow graph neural networks to process and analyze large-scale graph-structured data efficiently.

Start with graph neural networks by installing PyG, loading a small graph into a Data object, and running a minimal two-layer GCN to observe graph convolution. This quick start approach demonstrates core concepts and training patterns in action.