Error, Loss, and Metric

72 min
Block 8 — Training the model
Objective
separate three notions that the single English word "error" routinely collapses into one — the discrepancy on a single example, the differentiable objective the optimizer actually minimizes over the training set, and the non-differentiable business-facing quantity used to judge and select the model. Give each a rigorous definition, establish what statistical functional each loss estimates, explain why the loss and the metric are usually different and why that divergence is structural rather than a defect, and operate the scikit-learn scoring interface without falling into its sign convention traps.
Estimated duration
55 minutes
Prerequisites
chapters 008 (algorithm and model), 009 (parameters and hyperparameters), 026 (train/validation/test split) and 029 (fit, predict, predict_proba)
Associated quizzes
030.1-quiz-the-three-objects.md to 030.8-quiz-scoring-interface.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 “Training the Model” 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