Validation set
Definition
A slice of data kept out of training and used to check the model and tune its settings while you build it.
In plain English
The model never learns from the validation set, so it works like a practice exam: it shows how well the model handles data it hasn’t seen.
It is different from the test set. You use the validation set over and over while choosing settings, and save the test set for one final, honest score at the end.
Example
You split 10,000 labeled photos into 8,000 for training, 1,000 for validation, and 1,000 for the final test. You pick the best model based on the validation 1,000.
Draft, not yet reviewed by a person · Updated Nov 2026