Training Set
The portion of labeled data used to fit the model parameters.
- Typical split: 70% train, 15% validation, 15% test.
- Model sees these labels during fitting.
- Validation is for tuning; test is touched only at the end.
When to use: Always; the split is the first defense against fooling yourself about model quality.
Example: Take 100k labeled reviews: train 70k, validation 15k, test 15k. Fit on train, tune on validation, report final numbers from test only once.