Development version
This is the latest (dev) documentation. It may contain unreleased features or breaking changes. For the stable release, use stable.
Examples
Short, focused recipes — each example is self-contained and demonstrates a single feature you can copy directly into a script or notebook.
Examples vs Tutorials
Examples are focused recipes that each show one capability (a scorer, a plot, checkpointing, …). Tutorials are longer, end-to-end walkthroughs of a complete real-world task — from raw data to a tuned, evaluated model — usually integrating a specific library (XGBoost, LightGBM, CatBoost, …).
Hyperparameter Search
| Example | What it covers |
|---|---|
| Comparing Search Methods | Side-by-side: GASearchCV vs RandomizedSearchCV vs GridSearchCV |
| Advanced Random Forest Tuning | Smart initialization, warm starts, diversity control, fitness sharing, local search, adaptive schedules |
| Pipeline Regression | Pipeline parameter naming, regression scorers, search visualization |
Feature Selection
| Example | What it covers |
|---|---|
| Finding the Signal in 60 Columns | GAFeatureSelectionCV recovers the signal from a dataset that is two-thirds noise, beating the all-features baseline |
| Advanced RF + Feature Selection | Feature selection after hyperparameter tuning |
Multi-Metric and Refit
| Example | What it covers |
|---|---|
| Multi-Metric Search on Imbalanced Data | Multiple scorers that genuinely disagree, choosing the refit metric, inspecting per-metric cv_results_ |
Experiment Tracking
| Example | What it covers |
|---|---|
| MLflow 3 Experiment Tracking | Parent/child runs, dataset inputs, logged models, model lifecycle tags |
Visualization
| Example | What it covers |
|---|---|
| Plotting Gallery | plot_fitness_evolution, plot_history, plot_search_space |
Persistence
| Example | What it covers |
|---|---|
| Checkpointing and Persistence | ModelCheckpoint, save, load, inspecting checkpoint contents |