v0.8.0

version 0.8.0

Changed

Rework cdist_dtw and underlying Numba jited DTW computation to speed things up for numpy Backends.
Masks as returned by compute_mask, sakoe_chiba_mask and itakura_mask now have boolean values.
Backend attributes are now dynamic.
Ensure dtype is preserved in to_time_series_dataset
Fixed centroids computations in K-shape for multivariate timeseries (https://github.com/tslearn-team/tslearn/issues/288)
Fixed json and pickle serialization for LearningShapelet (https://github.com/tslearn-team/tslearn/issues/387)(https://github.com/tslearn-team/tslearn/issues/403)

Added

DBSCAN estimator for timeseries (https://github.com/tslearn-team/tslearn/issues/598)