Split it differently
Show on a big screenSame method, same crates. Move the split line and watch the score.
Loading lab…
- Same twenty-four crates. Move the line.
- Line at 6 marked: learn 10, test 14. Score 2 of 14.
- Line at 7 marked: learn 12, test 12. Score 7 of 12.
- Same method. Different test set. Different score.
The rule
The line is a number of marked fruit. Crates with fewer marked than the line are learned from. The others are kept back.
Move the split and measure again. The count and the number kept back are reported together. 3 of 4 and 30 of 40 are not the same claim.
The 40-crate test is ten times the size of the 4-crate test. So the two scores make different claims.
| Line at | Learn | Kept back | Score |
|---|---|---|---|
| 6 marked | 10 | 14 | 2 of 14 |
| 7 marked | 12 | 12 | 7 of 12 |
| 8 marked | 13 | 11 | 8 of 11 |
The method inside the program did not change. The line moved from 6 to 7. So the learn set grew by 2 crates. The count went from 2 to 7. That is more than 3 times as many.
At 6 marked, the learn set has no cut-price example. The program can only guess full. It misses every cut-price crate. At 8 marked, the learn set has two cut-price examples. The score is 8 of 11.
Watch out!
Do not report only the best split. Say which test set the number is for.
Try the lab to finish this lesson.