easy5 pts · first answer countsml-basicsevaluation

Why keep a separate test set

Why do ML teams keep a test set that is never used during training or tuning?

A.To make training run fasterB.To store extra data in case the training set is lostC.To get an honest estimate of how the model performs on new dataD.Because the model needs it to compute gradients

← Grouping data with no labels Spotting overfitting →