Holdout data Machine Learning
noun phrase
Definition: Examples intentionally withheld from training and used to assess how well a model generalizes to previously unseen data; validation and test datasets are both examples of holdout data [Google’s ML Glossary].
Example in context: “K-fold cross-validation [10] is an example of holdout data training and validation.” [Chen et al. 2024]
Synonyms: holdout set; hold-out set
Related terms: test set; validation set; train/test split