rodrigo-arenas/Sklearn-genetic-opt

Summary The search estimators expose cv_results_, but downstream users often want a stable, DataFrame-friendly way to inspect candidates, ranks, and parameters without manually unpacking param_* columns. Suggested scope Add a small public helper or estimator method to return candidate results as a pandas.DataFrame. Include parameter columns, mean/std test metrics, ranks, and optionally generation/index metadata when available. Preserve the existing cv_results_ contract; this should be an additive convenience API. Add tests for single-metric and multi-metric search results. Document the helper with a compact example. Acceptance criteria Users can inspect sorted candidate results with one documented call. The helper works for GASearchCV and GAFeatureSelectionCV where possible. No existing cv_results_ keys or shapes change. Pointers sklearn_genetic/genetic_search.py sklearn_genetic/utils/cv_scores.py sklearn_genetic/tests/test_genetic_search.py sklearn_genetic/tests/test_feature_sele

Summary Audit and improve checkpoint resume so it preserves the full optimization runtime state needed for reliable continuation. Why this helps Long-running searches need dependable resume behavior. Today the fit flow loads estimator_state and logbook from ModelCheckpoint, but it is worth auditing whether cache state, history, generation progress, random number state, and fit_stats_ are restored completely and consistently. Suggested implementation Start by reading: sklearn_genetic/genetic_search.py Review the checkpoint loading paths in both GASearchCV.fit and GAFeatureSelectionCV.fit. sklearn_genetic/callbacks/model_checkpoint.py Review what is saved today. sklearn_genetic/_base.py Review the generic save/load behavior. sklearn_genetic/tests/test_genetic_search.py Review existing checkpoint tests. Design a resume test that intentionally stops or checkpoints a short search, resumes it, and verifies that already evaluated candidates are not needlessly re-evaluated and th

Summary We would love to collect community-written tutorials, blog posts, notebooks, and articles about using sklearn-genetic-opt. Articles can live outside the project docs, for example on Medium, Dev.to, a personal blog, LinkedIn, GitHub Pages, or another public writing platform. After publishing, contributors can open a small pull request adding the link to the Community Articles page. Why this helps External articles help new users learn from real workflows and give contributors public recognition for teaching others. The project docs stay focused on official documentation, while this page makes community writing easier to discover. Suggested article topics Hyperparameter tuning with GASearchCV Feature selection with GAFeatureSelectionCV Using sklearn-genetic-opt with sklearn Pipelines Comparing genetic search with GridSearchCV or RandomizedSearchCV Practical tips for designing search spaces Real-world examples with imbalanced data, regression, or model selection Suggested implem

Summary Add lightweight smoke tests for a small subset of generated docs example scripts. The docs examples are valuable, but it is easy for imports or basic API usage to drift over time. A small smoke-test command would catch simple breakages without running every heavy example. Why this helps Examples are often the first code users copy. Smoke tests give maintainers confidence that selected docs examples still import, build their estimators, and run with deterministic small settings. Suggested implementation Start by reading these files: docs-vitepress/examples-src/ Look at the example source files and choose a few low-cost examples. docs-vitepress/examples-src/README.md Check how examples are generated. docs-vitepress/scripts/ Follow the style of existing docs scripts. Suggested approach: Add a small smoke-test entry point for docs examples. Start with only 2 or 3 low-cost examples. Avoid heavyweight optional dependency examples such as XGBoost, LightGBM, and CatBoost fo