Download Latest Version v0.9.0 source code.tar.gz (19.4 MB)
Email in envelope

Get an email when there's a new version of tslearn

Home / v0.9.0
Name Modified Size InfoDownloads / Week
Parent folder
README.md 2026-06-30 1.4 kB
v0.9.0 source code.tar.gz 2026-06-30 19.4 MB
v0.9.0 source code.zip 2026-06-30 19.5 MB
Totals: 3 Items   39.0 MB 0

v0.9.0

Added

  • Allow parallel computation of DTW barycenters and plug it in TimeSeriesKMeans.
  • NonMyopicEarlyClassifier can now be used with yet incomplete series or streamed inputs to retrieve optimal classification timing.
  • Add mae, mse and mase performance metrics for datasets
  • TimeSeriesFeatureSynchronizer is a new preprocessing to deal with acquisition of features at different sampling rates or desynchronized timestamps.
  • Add root_dir parameter to UCR_UEA_datasets to allow users to specify where downloaded datasets should be cached. If None, a default directory is used, either in $XDG_DATA_HOME/tslearn/UCR_UEA if the environment variable is set or ~/.tslearn/datasets otherwise.
  • Add VARIMA and AutoVARIMA forcasting estimator.

Changed

  • Fix masking issue in LearningShapelets with Keras 3.14
  • Fix imputer crash (#659)
  • TimeSeriesScalerMinMax and TimeSeriesScalerMeanVariance now scales based on fitted data characteristics when per_timeseries is False. (#280)
  • Fix LearningShapelets.shapelets_ attribute computation.
  • Fix LearningShapelets with MPS due to unsupported float64 precision
  • Rework dtw and frechet computation for torch backend to speed things up.
Source: README.md, updated 2026-06-30