All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
Changelogs for this project are recorded in this file since v0.2.0.
dtw_barycenter_averaging
is made faster by using vectorized computationsdtw_barycenter_averaging
can be restarted several times to reach better local optima using a parametern_init
set to 1 by default- Functions
load_timeseries_txt
andsave_timeseries_txt
from the utils module have changed their names toload_time_series_txt
andsave_time_series_txt
. Old names can still be used but considered deprecated and removed from the public API documentation for the sake of harmonization - Default value for the maximum number of iterations to train
ShapeletModel
andSerializableShapeletModel
is now set to 10,000 (used to be 100) TimeSeriesScalerMeanVariance
andTimeSeriesScalerMinMax
now ignore any NaNs when calling their respectivetransform
methods in order to better mirror scikit-learn's handling of missing data in preprocessing.
KNeighborsTimeSeriesRegressor
is a new regressor based on k-nearest-neighbors that accepts the same metrics asKNeighborsTimeSeriesClassifier
- A
set_weights
method is added to theShapeletModel
and
SerializableShapeletModel
estimators subsequence_path
andsubsequence_cost_matrix
are now part of the public API and properly documented as such with an example use case in which more than one path could be of interest (cf.plot_sdtw.py
)verbose
levels can be set for all functions / classes that usejoblib
for parallel computations andjoblib
levels are used;- conversion functions are provided in the
utils
module to interact with other Python time series packages (pyts
,sktime
,cesium
,seglearn
,tsfresh
,stumpy
,pyflux
) dtw_barycenter_averaging_subgradient
is now available to compute DTW barycenter based on subgradient descentdtw_limited_warping_length
is provided as a way to compute DTW under upper bound constraint on warping path length
- When constrained DTW is used, if the name of the constraint is not given but its parameter is set, that is now considered sufficient to identify the constraint.
- The
tests
subdirectory is now made a python package and hence included in wheels
- The way version number is retrieved in
setup.py
was not working properly on Python 3.4 (and made the install script fail), switched back to the previous version
- A
RuntimeWarning
is raised when an'itakura'
constraint is set that is unfeasible given the provided shapes.
'itakura'
and'sakoe_chiba'
were swapped inmetrics.compute_mask
tslearn
estimators are now automatically tested to matchsklearn
requirements "as much as possible" (cf.tslearn
needs in terms of data format, etc.)cdist_dtw
andcdist_gak
now have an_jobs
parameter to parallelize distance computations usingjoblib.Parallel
n_jobs
is also available as a prameter insilhouette_score
,TimeSeriesKMeans
,KNeighborsTimeSeries
,KNeighborsTimeSeriesClassifier
,TimeSeriesSVC
,TimeSeriesSVR
andGlobalAlignmentKernelKMeans
- Faster DTW computations using
numba
tslearn
estimators can be used in conjunction withsklearn
pipelines and cross-validation tools, even (for those concerned) with variable-length data- doctests have been reduced to those necessary for documentation purposes, the
other tests being moved to
tests/*.py
- The list of authors for the
tslearn
bibliographic reference has been updated to include Johann Faouzi and Gilles Van de Wiele - In
TimeSeriesScalerMinMax
,min
andmax
parameters are now deprecated in favor ofvalue_range
. Will be removed in v0.4 - In
TimeSeriesKMeans
andsilhouette_score
,'gamma_sdtw'
is now deprecated as a key formetric_params
in favor ofgamma
. Will be removed in v0.4
- Barycenter methods implemented as estimators are no longer provided: use
dedicated functions from the
tslearn.barycenters
module instead