Loading…
Glossary term
Data held back from training and used to tune settings and check progress.
The validation set is data held back during training and used to check progress and compare settings. If validation performance stops improving, that is a sign to stop or change something.
Because you make decisions based on it, it slowly becomes part of the tuning process, which is why a separate test set is kept for the end.
It lets you catch overfitting before it costs you, without touching the final test data.