- Faster C++ implementation
- Add
n_clusters
argument tobcn
to specify the number of clusters to be found (for now, by using k-means clustering) - Avoid division by zero in scaling the data
- Change criterion for early stopping to
utils::tail(abs(diff(errors_norm)), 1) >= tol
- Remove 'direct' method for now
- Improve
verbose
argument inbcn
(now an integer in (0, 1, 2, 3))
- Remove some dependencies to speed up package's startup
- Initial stable version