What is the purpose of a separate test set in addition to a validation set?
- A It gives an unbiased estimate because it is never used for tuning decisions
- B It doubles the training data
- C It speeds up training
- D It is required by scikit-learn
Answer
It gives an unbiased estimate because it is never used for tuning decisions
Every tuning decision made against the validation set leaks a little information, so validation performance becomes optimistic over many experiments.





