Connect input–transformation–evidence for school projects

Input — what examples enter (practice texts, filmed shots, quiz items). Include who labeled them and how you split held-out data.

Transformation — what the technique actually does (fits a flexible or simpler rule; compares scores; stops early). Say how quirk-memorization can happen.

Evidence — train vs validation (and maybe a friend’s set). Keep the worked numbers handy: flexible 99/84 (gap 15) vs simpler 93/91 (gap 2). State what is proved on your n, and what is unproved.

Glue with decision rules: measured baseline, reproducible versions, no tuning on the final test, honest posters.

Fill:

| Box | Your project | | --- | --- | | Input | … | | Transformation | … | | Evidence | … | | Failure you watch for | tuning only on practice answers | | Knob you might turn | simpler rules / stop early / more variety | | Claim you will not make | “works on all phones / all quizzes forever” |

Explain overfitting to a younger student in under two minutes using memorizing quiz answers vs learning the idea. Then show why a smaller practice score can win if the surprise quiz rises. If you need adult jargon to finish, rewrite until you do not.

Mastery for this topic is not a new algorithm — it is connection. Your school project should show the same spine as the chapter: inputs you can list, a transformation you can describe without magic, and evidence that includes a held-out or friend check. Stick the 99/84 vs 93/91 story next to your own numbers even if your scores differ; the pattern teaches the judgment.

Before you submit, read the card aloud. Cross out any sentence that could still be true for a project that only memorized practice answers. What remains should mention a surprise quiz, a gap, or a refused overclaim. That edit is the mastery gate. Re-evaluate when your data source, labeling friend, or project goal changes — yesterday’s evidence does not automatically cover today’s new inbox samples.