Every slice takes a turn as the judge
One validation split can be lucky or unlucky. Cross-validation gives every slice of the data a turn as the judge, then averages the scores. Scroll to panel B.
Try this
- Panel B: set k to 5, press Run cross-validation and watch the folds rotate.
- Press Pick best degree by CV, then reveal the test dots. Do it for three fresh datasets.
- Then press Repeat split 10× and compare the two rows of picks. The CV row should cluster tighter. That is the "better judge" moment.
This is part of Session 2 of How AI Learns. The full session adds the story, the Python build and a boss challenge.
Start free →