Preprint
Jul 2026
Hyperparameter Transfer in Graph Neural Networks
A transfer parameterization is developed and validated for GNNs trained with SGD, Adam, and AdamW that yields stable feature updates, learning rate transfer, and improved performance as width and depth increase, and provides a practical recipe for scaling GNNs across a variety of learning tasks and training scenarios.
G. Dezoort, Boris Hanin
· 0 citations