AI Blog

AI Blog

by Michele Laurelli

Validation Set

/ˌvælɪˈdeɪʃən sɛt/
Concept
Definition

A portion of data held out during training to tune hyperparameters and prevent overfitting.

The validation set is used to evaluate model performance during training and make decisions about hyperparameters, architecture, or early stopping. It's distinct from the test set which is only used for final evaluation.

Examples

1

80% train, 10% validation, 10% test split

2

Hyperparameter tuning

3

Early stopping criterion

Michele Laurelli - AI Research & Engineering