torch-geometric

Build a Graph Neural Network for node classification by loading graph datasets like Cora and implementing architectures such as GCN or GAT using PyTorch Geometric. This Skill provides scripts to train models for predicting node labels within graph-structured data.

Train models for molecular property prediction by representing chemical compounds as graph datasets and applying Graph Neural Networks. This Skill supports processing molecular structures to predict chemical properties using specialized GNN layers and transforms.

Supported graph neural network architectures for graph analysis include GCN, GAT, and GraphSAGE. This Skill enables implementing these models and custom layers to perform link prediction, node classification, and graph classification on diverse graph datasets.

PyTorch Geometric supports heterogeneous graphs and mini-batching for deep learning through specialized transforms and data handling utilities. This Skill facilitates utilizing these advanced features to scale graph neural network training across complex graph-structured data.

Use explainability tools with graph neural networks in PyTorch by applying integrated transforms and analysis functions. This Skill helps interpret GNN model predictions on graph-structured data, clarifying how node features and graph topology influence classification results.

Load custom graph datasets for graph neural network training by utilizing PyTorch Geometric's data handling and transform utilities. This Skill provides scripts to process and structure custom graph data, ensuring compatibility with GCN, GAT, and GraphSAGE model architectures.