Build a classifier from labeled data.
Separate model selection from final performance evaluation.
Choose two correct statements.
Training data learns parameters, validation data supports development choices, and a held-out test set is reserved for final evaluation without overlap.
Detailed explanation
Development and final evaluation remain separate.
The test is no longer unknown.
The test set becomes part of development.
The model learns patterns from these examples.
Check official documentation
Compare the following points with the official documentation.
Review Task 1.3 and the guidance on training, validation, and test separation. Map parameter fitting, model or setting selection, and the independent performance estimate after selection to their respective datasets.
Official references
What to confirm
Explain the role of each split and avoid repeatedly using test results to select models or settings, preserving an independent estimate.
Key points
- Training
- Validation
- Test
Notes
- Environment: Reference review of official documentation and hypothetical examples; no AWS resources or model training are required.
- Final evaluation describes the role of held-out data: it must not guide model or setting selection. Recomputing results for the same fixed model does not by itself establish leakage.
- If test results guide further tuning, that set has influenced development. A subsequent independent estimate needs other held-out data or an appropriate evaluation design.
Foundation review
Training data
Examples used to learn a model's parameters.
Validation data
Examples used during development to compare models or tune hyperparameters.