Model evaluation
Note
This section is in progress. There are no evaluation articles here yet — only this outline, plus empty Labs and Resources pages so the section has a home.
Why this section exists
Training a model is only half the work. The other half is asking, honestly: how well does it do, on data it has not seen, and by which measure?
What will live here
Pages I intend to write, not pages that exist today:
- Train, validation and test splits — why we hold data back.
- Metrics that match the problem (accuracy is not always the right number).
- Overfitting and underfitting, in pictures before formulas.
- Cross-validation, when the dataset is small.
- How evaluation changes when we move from classic models to deep networks and then to language models.
Nothing above is written yet. When a page appears, this outline will shrink and the sidebar will grow.