- Add classes_ attribute to fit method in BCNClassifier
- Add n_clusters parameter to fit method, to allow for a different number of clusters (for now, only k-means)
- Avoid division by zero when scaling the data
- Manage dependencies
- Return actual numpy arrays
- Remove 'direct' method for now
- Improve verbose (now an integer in (0, 1, 2, 3))
- Mirror the R package (remove some dependencies of R package).
- First release on PyPI.