Releases: VarIr/scikit-hubness
scikit-hubness 0.21.2
scikit-hubness 0.21.1
This is a bugfix release due to the recent release of scikit-learn 0.22.
For now, scikit-hubness specifically requires scikit-learn 0.21.3,
until the necessary adaptions to v0.22 are completed.
scikit-hubness 0.21.0
This is the first major release of scikit-hubness.
The package is organized as follows:
skhubness.analysis
provides theHubness
class, with which you can analyse your data for the presence and extent of hubness.
For example, you can assess the skewness of the k-occurrence distribution, or its Robin Hood index.skhubness.neighbors
provides exact and approximate nearest neighbor search, with or without hubness reduction. The package acts as a drop-in replacement forsklearn.neighbors
. You will find all its estimators, classes, and functions plus ANN and hubness reduction support (transparently using methods fromskhubness.reduction
).
For example, you can apply hubness-reduced kNN classification to large, high-dimensional data sets.skhubness.reduction
provides hubness reduction methods. The classes in this package can be used to transform kNN graphs directly.
For example, you can compute a mutual proximity kNN graph.
Find documentation at https://scikit-hubness.readthedocs.io/.
The package is available at https://pypi.org/project/scikit-hubness/ and can be installed with pip:
$ pip install scikit-hubness
All major platforms are supported.
Please have a look at the Changelog to see what's new compared to pre-releases.
scikit-hubness 0.21.0-alpha9
This is mostly a maintenance release with many smaller improvements.
Two more ANN libraries are now available on MacOS: NGT and Puffinn.
scikit-hubness 0.21.0-alpha8
Introduces new approximate nearest neighbor tools:
annoy
ngtpy
puffinn
scikit-hubness 0.21.0-alpha7
This release fixes several installation issues:
- including example dataset in wheel
- Specified versions of required packages (e.g.
scikit-learn>=0.21
) - added license notes
scikit-hubness 0.21.0-alpha4
The first alpha release of scikit-hubness
.