Validation Set
/ˌvælɪˈdeɪʃən sɛt/
A portion of data held out during training to tune hyperparameters and prevent overfitting.
In italiano: Validation SetThe 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
- 80% train, 10% validation, 10% test split
- Hyperparameter tuning
- Early stopping criterion