Ship and explain the tiny linear model

A shippable lab artifact includes: how to run it, the metric result (holdout absolute error; slope sign matches the hours/score trend), the failure you can still reproduce (training on the test row, or reporting R² without the holdout error), the security gate for publishing per-learner residuals that re-identify students, and a rollback note. The user decision it supports remains: estimate slope/intercept from study hours and score a held-out row honestly.

print({'artifact':'hours-score linear fit','proved':['holdout mae'],'unproved':['new semester'],'owner':'ml-lab'})

Expected evidence: tiny-model ship note. Store this beside the fixture version so scores remain meaningful after content changes in train-a-tiny-model.

State operating limits for the tiny linear model in plain language: fixture size, offline vs live dependencies, and what would require a new eval set. Shipping train-a-tiny-model is honest scoping, not maximal confidence language.

Fill this table in your notes for the tiny linear model:

• Proved on small hours/score table with one held-out pair: …
• Unproved beyond the fixture: …
• Metric that blocks release: holdout absolute error; slope sign matches the hours/score trend
• Failure still reproducible: training on the test row, or reporting R² without the holdout error
• Security gate: publishing per-learner residuals that re-identify students
• Rollback: …

Ship the narrative only when the unproved list is honest. Reviewers trust narrow claims that support estimate slope/intercept from study hours and score a held-out row honestly more than maximal language that collapses under the first production oddity.

Hand your ship note to a peer and ask them to recreate a proved/unproved ship note with rollback without watching you type. If they cannot, your evidence is still tribal knowledge. Tighten the run command and the metric line until a stranger can validate the tiny linear model against small hours/score table with one held-out pair.

At the mastery and shipping stage for train-a-tiny-model, the job is narrower than finishing a product demo. You are creating one progressive evidence piece about small hours/score table with one held-out pair that later pages inherit without redefining success. Keep that fixture small enough to inspect by hand, keep outputs copy-pasteable as text, and refuse to narrate this baseline as if it were a production SLA: predict mean train score for the holdout before fitting.

For this page specifically, success looks like a proved/unproved ship note with rollback while still centering the user decision to estimate slope/intercept from study hours and score a held-out row honestly. If you cannot point to a file, command, or assertion that proves that for the tiny linear model, stay on this page instead of advancing.

Overfitting glossary

Previous