14. What’s new¶

14.1. julearn 0.3.6.dev92 (2026-10-14)¶

Bugfixes¶

Fix a bug in which RandomizedSearchCV using uniform distributions specifications used an extended search space Fede Raimondo (#307)
Fix a bug in which CBPM with corr_sign="posneg" and no significant features reported two output feature names for a single output column Fede Raimondo (#315)
Fix a bug in which PipelineCreator and TargetPipelineCreator named every repeated step after the second one as <name>_1 instead of <name>_1, <name>_2, … Fede Raimondo (#315)

API Changes¶

Raise the minimum supported versions to scikit-learn 1.6.1, statsmodels 0.14.0 and pandas 2.2.2 Fede Raimondo (#316)

Features¶

Allow setting of context_func for joblib-htcondor by Fede Raimondo (#303)

Misc¶

Delay log message formatting by Fede Raimondo (#301)
Configure the project to support uv as a package manager and bump minimum Python version to 3.12 Fede Raimondo (#307)
Avoid generating and skipping non-applicable test cases and run tests in parallel with pytest-xdist Fede Raimondo (#315)
Improve CI: skip the tests when the library did not change, cancel superseded runs, actually run the nodeps tests, test with the minimum supported dependency versions and run the tests in parallel Fede Raimondo (#316)

Deprecations and Removals¶

Temporarily remove the deslib optional dependency and disable DynamicSelection, as the latest DESlib release is not compatible with scikit-learn >= 1.6 Fede Raimondo (#317)

14.2. Julearn 0.3.5 (2026-03-06)¶

Bugfixes¶

Fix parenthesis escaping in column names by Synchon Mandal (#226)
Fix StratifiedBootstrap behaviour to not leak test and train indices during splitting by Synchon Mandal (#254)
Fix a bug in which inspecting predictions from ContinuousStratifiedKFold would provide overlapping fold results by Fede Raimondo (#285)
Fix a bug in which the PipelineInspector will fail for pipelines with hyperparameter tuning by Fede Raimondo (#286)
Fix a bug in which inspecting predictions will result in NaNs if y had a custom index by Fede Raimondo (#288)
Change minimum scikit-learn version to 1.5.0 by Fede Raimondo. (#291)
Fix a bug in which OptunaSearchCV would fail when expecting the group parameter in the fit method by Fede Raimondo. (#297)
Enable and use metadata routing for hyperparameter tuning estimators by Fede Raimondo (#298)

API Changes¶

Remove test_size and train_size parameters and change default value of n_splits from 5 to 200 for StratifiedBootstrap by Synchon Mandal (#254)

Improved Documentation¶

Include documentation on how to use the joblib and joblib-htcondor library to parallelize computation by Fede Raimondo (#278)
Update citation in README by Synchon Mandal (#279)
Update joblib usage docs by Synchon Mandal (#282)
Add JuRSE Code Pick shield in README by Synchon Mandal (#283)

Enhancements¶

Avoid parallel calls in FilterColumns if not specified by the user in the config module by Fede Raimondo (#277)

Features¶

Add a new global config option "enable_auto_escape_parenthesis" to escape parenthesis in column names by Synchon Mandal (#226)

Misc¶

Regular repository maintenance by updating .pre-commit-config.yaml, replacing black with ruff-format, updating tool configs and GitHub Actions workflows and dependency version constraints, dropping support for Python 3.8 and adding support for Python 3.12 and 3.13 by Synchon Mandal (#284)
Update julearn.viz module to support param version 2.3.0 by Fede Raimondo (#297)
Update versions of dependencies in pyproject.toml to support pandas 3.X.X, numpy 2.X.X and scikit-learn 1.8.X by Fede Raimondo (#299)

14.3. Julearn 0.3.4 (2024-10-17)¶

Bugfixes¶

Fix usage of optuna distributions as hyperparameters in older versions of scikit-learn by Fede Raimondo. (#268)

Enhancements¶

Remove final model fit requirement for inspector to be returned by run_cross_validation() by Fede Raimondo. (#270)
Add run_fit() that implements a model fitting procedure with the same API as run_cross_validation() by Fede Raimondo. (#271)
Optimise wrapping of steps and models in the pipeline only when a subset of features is being used, by Fede Raimondo (#274)
Place the final model CV split at the beginning instead of the end of the CV iterator wrapper by Fede Raimondo (#275)
Change the internal logic of run_cross_validation() to optimise joblib calls by Fede Raimondo (#293)

14.4. Julearn 0.3.3 (2024-05-16)¶

Bugfixes¶

Fix OptunaSearchCV issue (https://github.com/optuna/optuna-integration/issues/118) with an internal implementation until the issue is fixed in optuna-integration by Fede Raimondo (#265)

Improved Documentation¶

Fixed a typo in cross validation by Dante Culaciati (#258)

14.5. Julearn 0.3.2 (2024-05-03)¶

Bugfixes¶

Fix a bug in which tuning hyperparemters while using a target transformer will not work as a result of target transformation detection by Fede Raimondo (#249)
Update bokeh api calls to remove warnings by Fede Raimondo (#255)

Improved Documentation¶

Update documentation on Hyperparameter Tuning by Fede Raimondo (#262)

Enhancements¶

Add BayesSearchCV to the list of available searchers as ‘bayes’ by Fede Raimondo (#260)
Refactor how hyperparmeters’ distributions are specified by Fede Raimondo (#262)

Features¶

Add OptunaSearchCV to the list of available searchers as optuna by Fede Raimondo (#262)

Misc¶

Add a new GitHub Action to mark stale issues by Synchon Mandal (#224)
Add support for scikit-learn 1.4.x by Synchon Mandal (#244)
Remove pytest-lazy-fixture and enable support for pytest >= 8.0.0 by Synchon Mandal (#251)
Add all as optional dependencies to install all functional dependencies by Fede Raimondo (#260)

14.6. Julearn 0.3.1 (2024-01-23)¶

Bugfixes¶

Add BaseEstimator to ConfoundRemoval for Target by Sami Hamdan (#151)
Skip wrapping scorers when target transformers can inverse transform and avoid logging filtered warnings by Fede Raimondo (#236)
Fix alternative parameter not being used in the corrected t-test Fede Raimondo (#245)
Fix hyperparameter not being set in PipelineCreator.add when an object is used as model, by Fede Raimondo (#247)

Improved Documentation¶

Fix cross-referencing links in documentation by Synchon Mandal (#168)
Standardize Sphinx section and header ordering format by Synchon Mandal (#234)
Improve documentation language, fix typos and code snippets by Synchon Mandal (#235)

Enhancements¶

Make CBPM use sum by default by Sami Hamdan (#170)

Features¶

Add support for multiple grids in hyperparameter tuning by Fede Raimondo (#47)

Misc¶

Adopt ruff as the only linter for the codebase by Synchon Mandal (#231)
Improve codespell support and fix typos in documentation by Synchon Mandal (#232)
Adopt pre-commit for adding and managing git pre-commit hooks by Synchon Mandal (#233)

14.7. Julearn 0.3.0 (2023-07-19)¶

The general API and behavior of Julearn has been changed to make it easier to define and use pipelines. The documentation has been updated to reflect these changes. Basic functionality from the previous API is still present. However, it might require setting a few more parameters in the run_cross_validation() function.

14.8. Julearn 0.2.5 (2022-07-21)¶

API Changes¶

Make API surrounding registering consistently use overwrite by Sami Hamdan
Inner cv needs to be provided using search_params. Deprecating cv in model_params by Sami Hamdan (#146)
Add n_jobs and verbose to run_cross_validation by Sami Hamdan

Bugfixes¶

Fix a hyperparameters setting issue where the parameter had an iterable of only one element by Sami Hamdan (#96)
Fix installations instruction for latest development version (add --pre) by Fede Raimondo
Fix target transformers that only normal transformers are wrapped by Sami Hamdan (#94)
Fix compatibility with new scikit-learn release 0.24 by Sami Hamdan (#108)
Fix compatibility with multiprocessing in scikit-learn by Sami Hamdan
Raise error message when columns in the dataframe are nos strings by Fede Raimondo (#77)
Fix not implemented bug for decision_function in ExtendedDataFramePipeline by Sami Hamdan (#135)
Fix Bug in the transformer wrapper implementation by Sami Hamdan (#122)
Fix Bug Target Transformer missing BaseEstimator by Sami Hamdan (#151)
Fix Bug of showing Warnings when using confound removal by Sami Hamdan (#152)
Fix Bug registered scorer not working in dictionary for scoring by Sami Hamdan

Improved Documentation¶

Add What’s new section in DOC to document changes by Fede Raimondo
Add information on updating the What’s new section before releasing by Fede Raimondo
Update docs to make it more uniform by Kaustubh Patil
Add example for tranform_until by Shammi More (#63)
Add documentation/example for parallelization by Sami Hamdan

Enhancements¶

Bump minimum python version to 3.7 by Fede Raimondo
Refactor scoring to allow for registering and callable scorers by Sami Hamdan
Update model_selection and add capabilities to register searchers by Sami Hamdan
Update default behavior of setting inner cv according to scikit-learn instead of using outer cv as default by Sami Hamdan
Add tests and more algorithms to DynamicSelection by Sami Hamdan and Shammi More

Features¶

Add user facing create_pipeline function by Sami Hamdan
Add CV schemes for stratifying continuous variables, useful for regression problems. Check ContinuousStratifiedKFold and RepeatedContinuousStratifiedKFold by Fede Raimondo and Shammi More
Add CBPM transformer by Sami Hamdan
Add register_model by Sami Hamdan (#105)