3.2.0

• Added note about gcc version on macOS
• Support feature_importances_ and dump_model() of RGF estimators

3.1.0

• FastRGF updated and bug about small sample weight was fixed. (Thanks @StrikerRUS!)
• RGF supports absolute error loss for regression.

3.0.0

From 3.0.0, FastRGF supports officialy, not alpha version

• Installation improvement
• Fix COO matrix

To 3.0.0, all contribution is done by @StrikerRUS. Many thanks!

2.3.0

• FastRGF parameter validation (Thank you for @StrikerRUS!).
• Change FastRGF parameter name like sklearn model (Thank you for @StrikerRUS!).
• FastRGF on the fly compilation.
• Add docker image usage.
• Fix view in Jupyter notebook (Thank you for @StrikerRUS!).
• Improve model input efficiency (Thank you for @StrikerRUS!).

2.2.0

• Support FastRGF as alpha version Please see https://github.com/fukatani/rgf_python/blob/master/FastRGF.rst
• Overall refactoring, Change directory configuration (Thank you for @StrikerRUS! )
• Enable to import by import rgf.RGFRegressor or other model. You can also use conventional import method import rgf.sklearn.RGFRegressor (Thank you for @StrikerRUS! )

2.1.2

reupdate to 2.1.2

2.1.1

• Add cleanup API

2.1.0

We are extremely pleased to announce the release of rgf_python 2.1.0, many thanks to the contributors and all users!

• Support serialization by pickle and joblib.
• Changed some error message and improve debuggability.

Fix encoding problem for installation

We fixed #65
Thanks @StrikerRUS and @abczhqiang!

2.0.2

Hot fix.