Release Notes
=============

sklearn-fab v1.1.0 for Python 3.9 (2022/11)
-------------------------------------------

### New Features

* This release supports Python 3.9.

### Other changes

* Support OS is updated to Red Hat Enterprise Linux 9.0.
* sklearn-fab now supports FAB version 1.6.6 or higher.


sklearn-fab v1.1.0 for Python 3.8 (2022/02)
-------------------------------------------

### New Features

* This release supports Python 3.8.

### Other changes

* Support OS is updated to Red Hat Enterprise Linux 7.9 / CentOS 7.9.
* sklearn-fab now supports FAB version 1.6.2 or higher.


sklearn-fab v1.1.0 (2020/03)
----------------------------

### New Features

* Model hotstart is now supported.

* sklearn-fab Migration Guide is released to assist migration from SAMPO/FAB.

### API Changes

* All sklearn-fab estimator's constructor are changed to accept FAB learner parameters.

* APIs for referencing model information are changed.
    - Estimator attribute `fic_` is added which has final fic.
    - Estimator attribute `prediction_formulas_` is changed to `comps_` and `comp_ids_`.
    - `SklearnFABTree.get_gate_functions()` return value dict is changed.

* Experimental estimators are removed.
    - Logit gate B-Spline Regression/Classification.

### Other changes

* Support OS is updated to Red Hat Enterprise Linux 7.7 / CentOS 7.7.
* sklearn-fab now supports FAB version 1.6.2 or higher.


sklearn-fab v0.2.0 (2019/12)
----------------------------

### New Features

* Add an estimator for Multi target Bernoulli gate Softmax Classification.

* Improved model visualization.
    - add some estimator attributes and methods for model visualization.
        - estimator.tree_
        - estimator.prediction_formula_
        - estimator.assign_comp()
        - estimator.calc_decision_function()
        - estimator.variance_  # Only regression estimators.

    - improved usability of FAB/HME model visualization tool.
        - `sklearn_fab.utils.export_gate_tree_dot()` returns pydot.Dot object,
            and you can now change feature names, labels in the gate tree easily.

### Experimental Features

* Add estimators for single target Logit gate B-Spline Regression/Classification.

### API Changes

* Estimator class names are changed.
    - `SklearnFABRegressor` is replaced to `SklearnFABBernGateLinearRegressor`.
    - `SklearnFABClassifier` is replaced to `SklearnFABBernGateLinearClassifier`.

* `sklearn_fab.utils.export_gate_tree_vis()` is replaced to
    `sklearn_fab.utils.export_gate_tree_dot()`.


sklearn-fab v0.1.1 (2019/09)
----------------------------
### New Features

* sklearn-fab now supports FAB version 1.5.0 or higher.

### Known issues
* Executing fit() on SklearnFABRegressor and SklearnFABClassifier does not return self.
  The estimator instance will not be returned after fit().


sklearn-fab v0.1.0 (2019/08)
----------------------------
### New Features

* sklearn-fab is available

    sklearn-fab has scikit-learn compatible estimators using the FAB/HME engine.
    Now, supports the following FAB/HME engine features:
      - Single target Bernoulli gate Linear Regression component Learner (Bern/Rg)
      - Single target Bernoulli gate Linear Classification component Learner (Bern/Cl)

*  FAB/HME model visualization

    Tool to visualize FAB/HME model (supervised binary tree models) is provided.

### Known issues
* Executing fit() on SklearnFABRegressor and SklearnFABClassifier does not return self.
  The estimator instance will not be returned after fit().