Now we have labeled pairs (x, y) and want a function f(x) ≈ y that works on new data too.

Regression vs. classification: regression means y is continuous (price, temperature); classification means y is a category (spam or not).