Why this day exists

Yesterday we chose a prediction rule from examples. Today we need to see one work on measured data.

Imagine an unfamiliar flower beside a tray of named flowers. You compare its measurements with the flowers you already know. Nearby examples can help suggest its name.

But checking only the tray you learned from would be an easy test. We need some flowers that stay outside the learning process.

A useful first model should be small enough to explain, and tested carefully enough that its score means something.