machine-learning
Domain inferred from 5 atoms across the corpora.
Atom counts by kind
| Kind | Count |
|---|---|
| anti-pattern | 1 |
| fact | 1 |
| pattern | 1 |
| principle | 1 |
| rule | 1 |
Sample atoms
Leaky Target
A feature that is computed using information that would not have been available at inference time — typically because the feature pipeline reads the latest value of some column instead of the value as-of the prediction t…
Evaluation Set Frozen
Generalization guarantees rely on the test set being statistically independent of the modeling decisions.…
Feature Store
A centralized system that defines, computes, stores, serves, and versions features for ML models — providing a single source of truth so the same feature definition is used in offline training and online inference.
Train Serve Skew
Train-serve skew is the gap between the feature distribution a model saw during training and the distribution it sees in production.…
Model Card Required
A Model Card is a 1-3 page structured document, version-controlled alongside the model artifact, containing nine sections: (1) Model Details — name, version, date, owner, license, citation; (2) Intended Use — primary use…