Clustergram
Bugfix for from_data method with non-default indices.
Bugs:
BUG: cluster centers empty due to index mismatch (#19)
Clustergram now supports interactive plotting using a new .bokeh() method based on BokehJS. It can be handy for exploration of larger and more complex clustergrams or those with significant outliers.
Enhancements:
ENH: support interactive bokeh plots (#14)
ENH: skip k=1 in K-Means implementations (#18)
documentation restructuring
Spring comes with native hierarchical clustering and the ability to create clustergam from a manual input.
ENH: support hierarchical clustering using scipy (#11)
ENH: from_data and from_centers methods (#12)
API chages:
pca_weighted is now keyword of Clustergram.plot() not init.
pca_weighted
Clustergram.plot()
Support MiniBatchKMeans (sklearn)
MiniBatchKMeans
Custom __repr__
Expose cluster labels obtained during the loop
Expose cluster centers
Silhouette score
Calinski and Harabasz score
Davies-Bouldin score
Version 0.2.0 brings support of Gaussian Mixture Models (using scikit-learn) and few minor changes.
Gaussian Mixture Model support (#4)
Verbosity - Clustergram now indicates the progress
Additional arguments can be passed to the PCA object
Bug fixes:
BUG: avoid LinAlgError: singular matrix