Knowledge Maxxing← Back to topic library
ML MaxxingBeginner60 sec
Validation set
The validation set is a portion of data held out from training, used to tune choices like hyperparameters and compare models during development.
Prerequisite: Training set
What matters
Because it is not used to fit the model's parameters, the validation set gives a more honest read of how choices generalize, before the test set is ever touched.
Example
A team might try three different model configurations and use validation-set accuracy to decide which one to move forward with.
Common misconception
The validation set is not the same as the test set; the validation set can be reused repeatedly during development, while the test set should not be.
Recall question
What is the validation set used for that the training set cannot be used for?
Keywords and related concepts
validation sethyperparameter tuningmodel selectiontraining set