.. _whatsnew_220:

.. currentmodule:: BPt

Version 2.2.0
---------------

- Initial support for Datasets concat
    - New :func:`concat` simmilar to :func:`pandas.concat` for
      concatenating with respect to :class:`Dataset` attributes.

- Improved inverse_transform_fis support
    - Added extra support for `inverse_transform_fis` method in :class:`BPtEvaluator`. 
    - Now supports some complex inverse transformations from nested :class:`Loader`, as well
      as some other bug fixes / new support.

- New support for `nested_model` in :class:`BPtEvaluator`
    - Methods `get_X_transform_df` and `permutation_importance` have a new parameter
      called `nested_model` which allows for transforming in nested pipelines

- :class:`Loader` bug fix
    - Call `fit_transform` on first data point instead of just `fit`.

- Updated compatibility with pandas 1.4
    - The update to pandas 1.4 was causing some issues with data types, these have been fixed.

- Removed RandomParcellation, SurfLabels and SurfMaps
    - Newer implementations, with more tests are available for these objects through the neurotools package.

- Add / changed some default pipelines
    - Changed name of svm pipe, and some new other default pipes

- Improving Test coverage
    - Added more misc tests, filling in gaps related to test coverage.