← Path
Machine Learning Fundamentals · Train, Validation and Test Sets: Understanding Overfitting
0 XP- 1 current stop: learning
- 2 upcoming
- 3 upcoming
- 4 upcoming
- 5 upcoming
Finish after stop 5
Stop 1 of 5 · Practice laps, race day, and the stopwatch
Trick to keep in mind+3 XP
Train fits it, validation picks it, test judges it once.
Parameters are fitted on training data, settings such as k are compared on validation data, and the test set gives one final unbiased score. If you keep choosing based on the test set, you are using it as a second validation set, and it stops being a fair judge. Training scores cannot choose a winner either, because memorisers always look great there.
Read it once, then a question tests whether the trick stuck.