Reducing overfitting

97 min
Block 9 — Generalization
Objective
turn the diagnosis of chapter 031 into a decision procedure. Establish that no remedy applies without a prior diagnosis, then work through the full catalogue of corrective actions — more data, better data, augmented data, fewer columns, less capacity, a penalty on the parameters, an interrupted optimization, an ensemble of decorrelated predictors, a smoothed target, a simpler problem — measuring the effect of each one rather than asserting it. Know the geometry that makes an L1 penalty produce exact zeros and an L2 penalty never produce them, read a cost-complexity pruning path, read a stopping trace, and price a remedy against the cost of buying more labeled rows. Close with a triage table that names, for each symptom, the cheapest remedy that works. Cover the reverse case as well: what to do when the model underfits, where every lever in this chapter must be pushed the other way.
Estimated duration
75 minutes
Prerequisites
chapters 025 (feature selection), 026 (train, validation and test), 031 (overfitting and underfitting) and 032 (bias-variance tradeoff)
Associated quizzes
033.1-quiz-diagnosis-first.md to 033.9-quiz-choosing-the-remedy.md

Preview — the rest of the lesson is for enrolled readers.

Already enrolled with a code?

Your access is tied to your account, not to this link. Sign in with the same email you used in class: your course is waiting, no need to enter the code again.

Sign inNo account yet? Create one
This lesson is part of the “Generalization, Overfitting and Underfitting” module

The first modules of the course are open to everyone. For the rest you have three options: buy this course once and for all, subscribe, or enter the code handed out in class.

Are you a student on this course?

The code is tied to your account: sign in or create an account and it will be applied automatically when you come back.

No account yet? Create one