Version 0.2.0

• Add this CHANGELOG file (#198)
• Update example of WHO feature selection (#212)
• Rename RidgeRegression2FoldCV -> Ridge2FoldCV (#211)
• Adding metrics for prediction rigidity (#209)
• Overhaul of documentation page (#200 to #204)
• Rename and add member variables (#197)
• Fix/check estimator (#196)
• fixed small typo in PCovR class documentation (#194)
• Resolved Issue WHO dataset missing function call section in doc (#181, #192)
• Speeding up tests (#190)
• Removing kernel optimization from who example (#189)
• Ignore rendered examples for linting (#188)
• Add more infos on documentation landing pages & CODE_OF_CONDUCT (#186)
• Add contributors pictures to README, show pip install instructions in docs (#185)
• Add linting and tests for docstring and documentation code (#184)
• Rerestructure requirements after (#171, #183)
• Update README.md to show banners (#176)
• Modernize package infrastructure (#172)
• Add an example of GCH for molecular materials (#171)
• Port examples to sphinx_gallery (#170)

Version 0.1.4

This is a patch release of skmatter that contains fixes for the documentation and directional convex hull

• documentation formatting fixes for math and datasets (#161 #163)
• changing the way the distance to the convex hull is computed in the DirectionalConvexHull due to numerical issues with the old method (#165)

Version 0.1.3

This is a patch release the last version of skcosmo before the refactor to skmatter

• Deprecation warning was added to link to renamed package (#154)
• dropped python <3.8 support, because we are now using scikit-learn version >=1.1.0 (#139 #146 #152 )
• WHO dataset and examples were added (#149)
• nice dataset was added (#143)
• overhaul of documentation (#142 #150)
• added DirectionalConvexHull class (#140)
• added test_precomputed_regression function to KPCovR (#136)
• other bugfixes (#141 #148)

Version skmatter-0.1.3

Refactor scikit-cosmo to scikit-matter #157 #151

Version 0.1.2

This is a patch release of skcosmo. It contains only bug fixes and small improvements, all users are encouraged to update.

• Allow to use a relative score threshold, which can detect identical points passed to FPS (#129)
• Update pi score less often in CUR (#133)
• Allowed saving more classes with pickle (#135)
• Allow selectors to select all available points (#137)
• other small bugfixes

Version 0.1.1

This is a patch release of skcosmo. It contains only bug fixes and small improvements, all users are encouraged to update.

• there was a bug in the orthonormalization step of PCov-CUR, which was fixed (#118)
• users can now initialize FPS selecting using a list of selected points, allowing to restart the selection in the middle (#116)
• KPCovR is now able to use pre-fitted regressor in the same way that PCovR can (#113)

Version 0.1.0

This is the first public release of scikit-cosmo. Scikit-cosmo is a collection of scikit-learn compatible utilities that implement methods developed in the COSMO laboratory. This first release contains multiple tools of general interest:

• Principal Covariate Regression (PCovR) and the kernel extension KPCovR
  • https://doi.org/10.1088/2632-2153/aba9ef
• Feature and sample selection methods: Farthest point sampling and CUR selection, as well as PCovR version of these methods
  • https://arxiv.org/abs/2012.12253
• tools to compare different features spaces: Global Feature Reconstruction Error (GFRE), Global Feature Reconstruction Distortion (GFRD), Local Feature Reconstruction Error (LFRE)
  • http://doi.org/10.1088/2632-2153/abdaf7

Have a look at the documentation if you are interested!