Trainer scripts and metrics for GNN training

ELEKTRONN / elektronn3

A PyTorch-based library for working with 3D and 2D convolutional neural networks, with focus on semantic segmentation of volumetric biomedical image data

MIT License

161 stars 27 forks source link

Trainer scripts and metrics for GNN training #47

Closed hash-ir closed 2 years ago

hash-ir commented 2 years ago

The following contributions are made:

Mini-batch GNN trainer: supports semi-supervised and self-supervised training in mini-batch mode
Full-batch GNN trainer: supports semi-supervised training in full-batch mode
Clustering metrics: Added mutual information scores, silhouette score, and adjusted rand index from scikit-learn