Learning Curves: More Data or More Model?

Classical models built from the math you already know

More data, or a bigger model? Money and time only allow one, so guessing wrong is an expensive mistake. Telling the two apart means watching two numbers as the training set itself grows: training error, and held-out error (often called validation error in practice), the score on data kept back and never shown to the fit. Plot both against training-set size and a genuinely useful diagnostic appears, called a learning curve.

🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.