Validation Data
Validation data is a held-back set of examples used during development to tune a model's settings and detect overfitting, separate from the data it is trained on.
What is Validation Data?
The EU AI Act describes validation data as data used to evaluate a trained system and tune its non-learnable parameters and learning process. Because developers make choices based on validation results, a further untouched test set is still needed for a final independent check.
How it works
After each round of training the model is scored on the validation set; the settings or checkpoint that perform best are kept.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions