Glossary term
Glossary term
Foundations
A metric representing a model's loss on the validation set during a particular iteration of training.
See also generalization curve.
Created for this library
An ML team uses validation loss to choose between model variants in a sweep before promoting any single candidate.
A research team logs validation loss every epoch to decide when to stop training and pick the best checkpoint.
An ML platform team includes validation loss alongside training and test metrics in every release review.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License