Version v0.9.1

What's Changed

New features

• feat: support polars lazyframe in add_lags by @MarcoGorelli in #675
• feat: ZeroInflatedRegressor.score_samples(...) by @FBruzzesi in #680

Fixes

• patch: .drop(columns=...) test hotfix by @FBruzzesi in #678
• patch: ci/cd hotfix by @FBruzzesi in #683
• patch: strict=False in polars constructor by @FBruzzesi in #684
• patch: FairClassifier regularization and fit intercept by @FBruzzesi in #679

Others

• perf: make sure X.schema is only calculated once per dataframe in TypeSelector by @MarcoGorelli in #676
• chore: use Narwhals stable api by @MarcoGorelli in #685
• version 0.9.1 by @koaning in #686

Full Changelog: v0.9.0...v0.9.1

Version v0.9.0

What's Changed

Disclaimer

This is a large one, with a lot of improvements, new features and bug fixes!

TL;DR

• narwhals support for dataframe agnostic codebase
• numpy 2.0 support
• cvxpy 1.5 support
• improvements in scikit-learn intergration

New Features ✨

• Narwhals for dataframe-agnostic codebase by @FBruzzesi in #671
  This was a very large effort with the help of many to introduce narwhals:

  • #655 by @MarcoGorelli
  • #659 by @anopsy
  • #661 by @MarcoGorelli
  • #665 by @MarcoGorelli
  • #667 by @FBruzzesi
  • #669 by @DeaMariaLeon
  • #670 by @MarcoGorelli

• numpy 2.0rc support by @FBruzzesi in #651
  Supports of numpy 2.0 for base library (not including cvxpy and umap)

Bug fixes 🐞

• cvxpy 1.5.0 support by @FBruzzesi in #663
• ZeroInflatedRegressor fix training indices by @FBruzzesi in #666

Internals ⚒️

• Improving sklearn compatibility via parametrize_with_checks by @FBruzzesi in #660

Full Changelog: v0.8.2...v0.9.0

Version v0.8.2

What's Changed

• Support for python 3.12 by @FBruzzesi in #628
• Fix broken links on home page by @iuliaferoli in #634
• Fix typos by @k-moun in #638, #640, #641
• Docstring example for OutlierRemover by @anopsy in #639
• Add bayesian methods to GMM density page by @mkalimeri in #642
• Adopt UV in GHA by @FBruzzesi in #629
• Docs rendering by @FBruzzesi in #645
• ClusterFoldValidation now with correct spelling by @koaning in #636
• Docstring example for DictMapper and OutlierClassifier by @anopsy in #646
• Contributing docs section by @FBruzzesi in #644
• Update this.py by @FBruzzesi in #647
• ClusterFoldValidation docs api section by @FBruzzesi in #649
• Docstrings example for FormulaicTransformer, IdentityTransformer, InformationFilter, PandasTypeSelector and RepeatingBasisFunction by @anopsy in #648

New Contributors

• @iuliaferoli made their first contribution in #634
• @k-moun made their first contribution in #638
• @anopsy made their first contribution in #639
• @mkalimeri made their first contribution in #642

Full Changelog: v0.8.0...v0.8.2

0.8.0

Description

✨ New Features

• GroupedClassifier and GroupedRegressor #619 by @FBruzzesi
• MaximumRelevanceMinimumRedundancy feature selector #622 by @fabioscantamburlo
• HierarchicalPredictor, HierarchicalClassifier and HierarchicalRegressor #623 by @FBruzzesi

🐞 Bug Fix

• GroupedTransformer allows transformers that use y #624 by @FBruzzesi
• Raise early for GroupedPredictor with shrinkage in non-regression tasks #619 by @FBruzzesi

⚠️ Breaking Changes

• Removed deprecated classes: PatsyTransformer, FairClassifier, OutlierRemover #626 and GroupedEstimator #619

0.7.4

Added a new meta estimator for ordinal classification.

0.7.0

@FBruzzesi created a new OrdinalClassifier meta model. While somewhat experimental ... we can totally see how it may be nice for some folks. Enjoy!

0.7.0

Thanks to @FBruzzesi we now have new docs pages and an update to the DecayEstimator. This estimator is now more flexible, but does introduce a breaking change. Hence a big number update!

v0.6.16

Now that @FBruzzesi is helping out with the maintenance we've got a bunch of PRs to celebrate! Some of these are related to relaxing the check_X_y function in some of the meta models.

• #584
• #582
• #580
• #577

v0.6.15

Mainly a patch release for #562.

0.6.14

Added support for GroupTimeSeriesSplit in #540.