Runnable examples¶
Choose a question before choosing a component. Every Python block in the new guides is standalone and executed during validation; examples use synthetic data or local arrays and need no external downloads.
| Question | Example | What to inspect |
|---|---|---|
| How wide is uncertainty for a numeric prediction? | Getting started | Test coverage and mean width |
| Does an interval cross a decision boundary? | Regression decisions | Review count and interval score |
| Which labels remain plausible? | Classification review | Class mapping, coverage and set size |
| What do intervals and sets look like? | Visual examples | Held-out intervals and prediction set sizes |
| How should I compare coverage and efficiency? | Diagnostics | Interval tail misses, interval scores, set sizes and class-wise coverage |
| Can I develop models with sklearn tools? | Cross-validation | Clone lifecycle and point scores |
| Where are outcomes missed? | Plotting | Predicted ranges and observed test targets |
The original workflows remain available, including supported outer-pipeline examples and their statistical caveats. Run examples with the locked environment described in development; keep calibration independent of all learning.