v1.4.0

March 2026

Highlights

• New native implementations of estimators, including: ROCKAD anomaly detector; K-Shape clusterer; TS-QUAD configuration for in ElasticEnsemble classifier; ETS/AutoETS and DeepARF forecasters; and STL, MSTL and LOWESS transformers

• Most classification and regression data is now loaded from Zenodo, including from the Multiverse, a relaunched multivariate classification archive

• Deprecate Python 3.10 and allow Python 3.14, along with altering version bounds for many core derpendencies

• Plenty of bug fixes and deprecation actions

Thanks to our 28 contributors with 89 pull requests since the last release.

Anomaly Detection

• [BUG] Fix broken fit_predict in anomaly detection & add tests (#3185) @Adityakushwaha2006

• [ENH] Add whole-series ROCKAD anomaly detector (#2871) @pattplatt

• [ENH] Improve and add testing to label validation (#3310) @MatthewMiddlehurst

Classification

• [BUG] Fix bug in WEASELTransformerV2.fit_transform() method (Issue #3283) (#3284) @patrickzib

• [BUG] Return TDE and HC2 to expected results test skip (#3322) @TonyBagnall

• [BUG] Enforce at least one of each class for dummy classifier test (#3341) @TonyBagnall

• [BUG] Enforce balanced test data for sklearn compatible testing (#3342) @TonyBagnall

• [DOC] Add missing exceptions to KNeighborsTimeSeriesClassifier Raises section (#3255) @Varshinibhargav-17

• [ENH] Expose n_shapelets_ attribute in RDST Transformer and Classifier (#3196) @satwiksps

• [ENH] Implement TS-QUAD configuration option in ElasticEnsemble (#3126) @Nithurshen

• [ENH] adding type hints for files in aeon/classification/deep_learning (#3241) @stephanielees

• [ENH] add unequal-length time series support for tsfresh-based methods (#3187) @jsquaredosquared

• [ENH] Improve and add testing to label validation (#3310) @MatthewMiddlehurst

• [DOC]: correct load_classification function name from 2023 -> 2021 (#3346) @kennaruk

Clustering

• [BUG] FIX DRNN load bug (#3074) @PipaFlores

• [BUG] KASBA bug fixes (#3120) @chrisholder

• [BUG] Petitjean averaging parallel flag (#3123) @chrisholder

• [DOC]: fix typo in custom_alignment_path docstring (#3243) @devesh-047

• [ENH] Validation loss for deep learning clustering #3039 (#3046) @PipaFlores

• [ENH] add unequal-length time series support for tsfresh-based methods (#3187) @jsquaredosquared

• [ENH] Port in kshape (#3314) @TonyBagnall

• [ENH] deprecate TSLearnKShape (#3343) @TonyBagnall

• [ENH] Cluster initialisation refactor (#3122) @chrisholder

• [MNT] Update tslearn dependency to >=0.7.0 and fix kshape tests (#3119) @Nithurshen

• [MNT] Add missing load_model test for deep clusterers (Fixes #3080) (#3111) @satwiksps

• [MNT] Update expected results for tslearn kshape (#3305) @MatthewMiddlehurst

• [BUG] MSM bug fixes (#3121) @chrisholder

Datasets

• [BUG] Add load_regression encoding parameter (#3087) @Cyril-Meyer

• [BUG] Fix imports for .ts saver and defaults for unequal transforms (#3089) @MatthewMiddlehurst

• [ENH] Add new datasets to monash list. (#3118) @alexbanwell1

• [BUG] Fix #3203 - Use proper file names in load_regression method (#3204) @Lucky-Lodhi2004

• [ENH] Update download_all_regression to use extended TSER 2024 archive (#3240) @devesh-047

• [ENH] Add loaders for Monster datasets with hugging_face dependency (#3141) @rwtarpit

• [ENH] Multiverse: change all loaders to zenodo, include new problems (#3304) @TonyBagnall

• [DEP] deprecate loader arguments (#3355) @TonyBagnall

• [DOC] Tidy up TSC lists (#3354) @TonyBagnall

Distances

• [BUG] MSM bug fixes (#3121) @chrisholder

• [ENH] Implement O(N) space complexity for dtw_distance (#3258) @raphaelgimenezneto

• [DOC] dtw_distance window parameter doc (#3200) @Lucky-Lodhi2004

Forecasting

• [BUG] Fix ETS bug (#3286) @alexbanwell1

• [ENH] ETS/AutoETS docs (#3295) @TonyBagnall

• [ENH] Added DeepARForecaster to aeon/toolkit (#3051) @lucifer4073

• [ENH] Implement series_to_series forecaster mixin (#3146) @rwtarpit

• [ENH] Add Working AutoETS algorithm (#3253) @alexbanwell1

• [ENH] Make the ACF function more efficient (#3296) @TonyBagnall

• [ENH] Add exogenous variable support to ARIMA (#3116) @satwiksps

Networks

• [BUG] Cast numpy scalars to int for Dense layer units (#3182) @Adityakushwaha2006

• [ENH] Added DeepARForecaster to aeon/toolkit (#3051) @lucifer4073

• [ENH] Validation loss for deep learning clustering #3039 (#3046) @PipaFlores

• [BUG] Initialize _metrics in build_model for all Deep Learning estimators (#3198) @satwiksps

• [ENH] Added NBeatsNetwork to aeon/networks (#3057)) @lucifer4073

Regression

• [ENH] add unequal-length time series support for tsfresh-based methods (#3187) @jsquaredosquared

• [ENH] Improve and add testing to label validation (#3310) @MatthewMiddlehurst

• [ENH] Add Feature load_model to Deep Regressor Ensembles (#3130) @rwtarpit

• [DOC] Add Raises section to KNeighborsTimeSeriesRegressor docstring (#3337) @Varshinibhargav-17

Segmentation

• [BUG] Fix zero-distance instability in Hidalgo (#3068) (#3115) @satwiksps

• [BUG] Fix HidalgoSegmenter IndexError with empty sampling (Fixes #3168) (#3169) @samay2504

• [ENH] Add primitive type hints to segmentation module (#3252) @vickysharma-prog

Transformations

• [BUG] Fix imports for .ts saver and defaults for unequal transforms (#3089) @MatthewMiddlehurst

• [BUG] Fix bug in WEASELTransformerV2.fit_transform() method (Issue #3283) (#3284) @patrickzib

• [ENH] Expose n_shapelets_ attribute in RDST Transformer and Classifier (#3196) @satwiksps

• [ENH] Make transformers inherit TransformerMixin and add CollectionToSeriesWrapper (#2812) @MatthewMiddlehurst

• [ENH] add unequal-length time series support for tsfresh-based methods (#3187) @jsquaredosquared

• [ENH] Optimize ROCKET GPU PPV calculation for 1.45× speedup (#3232) @Adityakushwaha2006

• [ENH] STL and MSTL transformer (#2960) @TinaJin0228

• [ENH] Separate inverse_transform as a mixin for transformers (#3018) @MatthewMiddlehurst

• [ENH] LOWESS smoother (#3321) @TonyBagnall

Visualisations

• [ENH] MCM plot function tidy. (#3099) @MatthewMiddlehurst

• [ENH] text overrun on plot_critical_difference #3098 (#3138) @Lucky-Lodhi2004

• [ENH] Temporal importance curves with sklearn decision trees (#3327) @MatthewMiddlehurst

Other

Bugs

• [BUG] Handling series with extremely small scale (#3315) @MatthewMiddlehurst

Deprecation

• [DEP] Remove deprecated data savers, validation methods and transform update method (#3309) @MatthewMiddlehurst

Documentation

• [Doc] A fix to inconsistent ticks in Base module (#2734) @Ahmed-Zahran02

• [DOC] Fix 15 broken estimator links by refactoring path resolution logic (#3172) @Adityakushwaha2006

• [DOC] Update community support channels to include Discord (#3129) @hadifawaz1999

• [DOC] Discord links and small workflow fix (#3256) @MatthewMiddlehurst

• [DOC] Fix typo in contributing guide (#3281) @NikVince

• [DOC] Add examples to validation utility functions (#3225) @nimra06

• [DOC] Add some AI policy documents and citation file (#3251) @MatthewMiddlehurst

• [DOC] Revise aeon paper citation and hyperlink (#3303) @MatthewMiddlehurst

• [DOC] Remove links to X/Twitter (#3317) @MatthewMiddlehurst

Enhancements

• [ENH] Add type hints to get_unchanged_and_required_params_as_str (#3229) @nimra06

• [ENH] Add n_timepoints in the metadata stored by BaseSeriesEstimator (#3163) @satwiksps

• [ENH] Record fit time in class attributes (#3331) @MatthewMiddlehurst

Maintenance

• [MNT] Update slack link (#3103) @MatthewMiddlehurst

• [MNT] Restrict tslearn version range in pyproject.toml (#3107) @hadifawaz1999

• [MNT] Replace platform-dependent np.int_ with np.intp to prevent Windows overflows (#3202) @Adityakushwaha2006

• [MNT] Install aeon all_extras for expected results action (#3094) @MatthewMiddlehurst

• [MNT] Update numba versioning (#3246) @MatthewMiddlehurst

• [MNT] numpy and scipy upper bound updates (#3247) @MatthewMiddlehurst

• [MNT] Change stale branch check to run monthly, wait longer and stop tagging people (#3280) @MatthewMiddlehurst

• [MNT] Parallelize Example Notebook Runs in CI (#3162) @satwiksps

• [MNT] Update MANIFEST.in (#3298) @MatthewMiddlehurst

• [MNT] Deprecate Python 3.10 and allow Python 3.14 (#3245) @MatthewMiddlehurst

• [MNT] raise scikit requirement to 1.6 or higher (#3340) @TonyBagnall

• [BUG] Enforce balanced test data for sklearn compatible testing (#3342) @TonyBagnall

• [MNT] Check HC2 and TDE expected results testing (#3307) @MatthewMiddlehurst

• [MNT] Skip tsfresh multithreading test (#3325) @MatthewMiddlehurst

• [REF] Refactor numba thread decorator (#3332) @MatthewMiddlehurst

• [MNT] Update dependencies and configuration in pyproject.toml (#3167) @TUSHARDHOKRIYA

• [DOC] Fix typos and enhance clarity in getting_started.md (#3134) @PoojasPatel013

Contributors

The following have contributed to this release through a collective 89 GitHub Pull Requests:

@Adityakushwaha2006, @Ahmed-Zahran02, @alexbanwell1, @chrisholder, @Cyril-Meyer, @devesh-047, @hadifawaz1999, @jsquaredosquared, @kennaruk, @lucifer4073, @Lucky-Lodhi2004, @MatthewMiddlehurst, @NikVince, @nimra06, @Nithurshen, @patrickzib, @pattplatt, @PipaFlores, @PoojasPatel013, @raphaelgimenezneto, @rwtarpit, @samay2504, @satwiksps, @stephanielees, @TinaJin0228, @TonyBagnall, @TUSHARDHOKRIYA, @Varshinibhargav-17, @vickysharma-prog