First steps towards v0.30

[VarIr]

scikit-hubness v0.30 will introduce support for recent scikit-learn versions. Note that this is not yet fully functional and may contain any number of missing features and bugs.

scikit-hubness v0.30 tackles the Outlook section from our JOSS paper:

"Future plans include adaption to significant changes of sklearn.neighbors introduced in version 0.22 in December 2019: The KNeighborsTransformer and RadiusNeighborsTransformer transform data into sparse neighbor graphs, which can subsequently be used as input to other estimators. Hubness reduction and approximate search can then be implemented as Transformers. This provides the means to turn skhubness.neighbors from a drop- in replacement of sklearn.neighbors into a scikit-learn plugin, which will (1) accelerate development, (2) simplify addition of new hubness reduction and approximate search methods, and (3) facilitate more flexible usage."

[lgtm-com[bot]]

This pull request introduces 15 alerts when merging b2f1a85c1cbbbd656889c82f2d5a9b2789ff23f8 into 6eaeedda2c4b52bb7bf2553b3c5b04a076287ae3 - view on LGTM.com

new alerts:

• 5 for Conflicting attributes in base classes
• 5 for Wrong number of arguments in a class instantiation
• 3 for Unused import
• 2 for Implicit string concatenation in a list