v1.2.0

June 2025

Highlights

• aeon v1.2.0 drops support for Python 3.9, Python 3.10-3.13 are the currently supported versions.

• Framework for whole-series anomaly detection has been implemented, and the module has been refactored into anomaly_detection.collection and anomaly_detection.series submodules.

• ProximityForest and ProximityTree classifiers have been improved to support unequal length and multivariate series. The classifiers are also significantly faster.

• A new implementation for the TD-MVDC classifier

• The forecasting module has been enhanced with additional methods for direct and recursive forecasting, as well as new algorithms such as ETS. This module is still in early development, and we expect framework changes in future releases.

• The similarity search module has been significantly reworked to fit the aeon style framework seen in other modules. See the module examples and documentation for more details.

• The start of an experimental sub-package for self supervised learning has been added in aeon.transformations.collection.self_supervised, starting with the TRILITE algorithm.

Anomaly Detection

Deprecation

• [DEP] Remove prts dependency (#2863) @MatthewMiddlehurst

Documentation

• [DOC] resolved the inconsistency of double ticks for the anomaly detection module #809 (#2546) @AnaghDeshpande

Enhancements

• [ENH] Refactor range-based metrics to restore original behavior (#2781) @SebastianSchmidl

• [ENH] add informational tags back to estimator docs (#2652) @inclinedadarsh

• [ENH] Update _continuous.py functions (#2822) @aryanpola

• [ENH] Whole-series anomaly detection (#2326) @MatthewMiddlehurst

• [ENH, REF] Refactored time-point based ROCKAD implementation (#2804) @pattplatt

Refactored

• [ENH, REF] Refactored time-point based ROCKAD implementation (#2804) @pattplatt

• [REF] Refactor Anomaly Detection Module into Submodules by Algorithm Family (#2694) @Kaustbh

Benchmarking

Bug Fixes

• [BUG] Switch results loaders from http to https (#2883) @TonyBagnall

Enhancements

• [ENH] Refactor range-based metrics to restore original behavior (#2781) @SebastianSchmidl

• [ENH] Update _continuous.py functions (#2822) @aryanpola

Classification

Bug Fixes

• [BUG] Better error message for Rotation Forest with no usable attributes (#2798) @MatthewMiddlehurst

Documentation

• [DOC] Add Raises section for invalid weights in KNeighborsTimeSeriesClassifier (#1766) (#2764) @SomtoOnyekwelu

• [DOC] A fix to inconsistent double tick quote for the classification.distance_based module (#2627) @Ahmed-Zahran02

• [DOC] Add RotationForest Classifier Notebook for Time Series Classification (#2592) @Kaustbh

Enhancements

• [ENH] Implement load_model function for ensemble classifiers (#2631) @inclinedadarsh

• [ENH] Use n_jobs parameter in KNeighborsTimeSeriesClassifier. (#2687) @steenrotsman

• [ENH] Implemented Tracking differentiator-based multiview dilated characteristics (TD-MVDC) Classifier (#2535) @lucifer4073

• [ENH] Multivariate and unequal PF, speed-ups and distance classifier tidy (#2872) @MatthewMiddlehurst

Clustering

Bug Fixes

• [BUG] Fix indexing errors in kdtw of KernelKMeans (#2826) @SebastianSchmidl

Documentation

• [DOC] Changed SummaryClusterer’s summary_stat from “percentiles” to “quantiles” (#2778) @Ramana-Raja

Enhancements

• [ENH] Fixes Issue Improve _check_params method in kmeans.py and kmedoids.py (#2682) @tanishy7777

• [ENH] Added test cases for feature based clustering (#2690) @Ramana-Raja

Datasets

Bug Fixes

• [BUG] Fix data loader (#2810) @TonyBagnall

• [ENH] Refactor example dataset for unequal length univariate (#2859) @TonyBagnall

Enhancements

• [ENH] Refactor example dataset for unequal length univariate (#2859) @TonyBagnall

Distances

Bug Fixes

• [BUG] Fix indexing errors in kdtw of KernelKMeans (#2826) @SebastianSchmidl

Enhancements

• [ENH] Multivariate and unequal PF, speed-ups and distance classifier tidy (#2872) @MatthewMiddlehurst

Forecasting

Bug Fixes

• [BUG] Forecasting regressor fix and notebook (#2885) @TonyBagnall

Documentation

• [DOC] Regression forecaster docstring (#2837) @TonyBagnall

Enhancements

• [ENH] Type hints/forecasting (#2737) @Ahmed-Zahran02

• [ENH] refactor dummy forecaster to naive (#2831) @TonyBagnall

• [ENH] introduce revised version of ETS (#2834) @TonyBagnall

• [ENH] Add two forecasting tags (#2882) @TonyBagnall

• [ENH] Forecasting testing (#2891) @MatthewMiddlehurst

• [ENH] Add recursive and direct methods to forecasting base class (#2899) @TonyBagnall

• [ENH] remove horizon from ETS constructor, parametrize ETS tests (#2898) @TonyBagnall

• [ENH] enhance naive forecaster with new strategies (#2869) @TinaJin0228

Networks

Bug Fixes

• [BUG] Fix bug of shape variable assignment in AEAttentionBiGRUNetwork (#2771) @hadifawaz1999

Enhancements

• [ENH] Adapt the DCNN Networks to use Weight Norm Wrappers (#2628) @aadya940

• [ENH] Test coverage for AEAttentionBiGRUNetwork Improved (#2540) @lucifer4073

• [ENH] Added RNN in networks (#2875) @lucifer4073

Regression

Bug Fixes

• [BUG] Better error message for Rotation Forest with no usable attributes (#2798) @MatthewMiddlehurst

Documentation

• [DOC] Docstring improved for dummy regressor (#2839) @lucifer4073

Enhancements

• [ENH] Add type hints for deep learning regression classes (#2644) @saadaltohamy

Segmentation

Bug Fixes

• [BUG] Fix a bug in ClaSP Series Transformer when using floatXX as Input (where XX != 64) (#2774) @patrickzib

Documentation

• [DOC] Inconsistent double ticks in segmentation package, public docstring only (#2622) @nMaax

• [DOC] Improved Hidalgo Segmentation Notebook (#2616) @kavya-r30

Maintenance

• [MNT] stop segmenters changing state in predict (#2526) @TonyBagnall

Similarity Search

Documentation

• [MNT, ENH, DOC] Rework similarity search (#2473) @baraline

• [DOC] Fix inconsistent double backticks in similarity_search module (#809) (#2618) @adityagh006

Enhancements

• [MNT, ENH, DOC] Rework similarity search (#2473) @baraline

Maintenance

• [MNT, ENH, DOC] Rework similarity search (#2473) @baraline

Transformations

Bug Fixes

• [BUG] Remove typing from numba functions (#2858) @TonyBagnall

• [BUG] Fixes a bug with SFAFast throwing an error when calling transform after fit (#2897) @patrickzib

Enhancements

• [ENH] Optimize QUANTTransformer by using shape calculation in _fit method to avoid unnecessary computations (#2727) @shinymack

• [ENH] Add missing checks to base transformers and tidy (#2773) @MatthewMiddlehurst

• [ENH] HydraTransformer Output changed (#2656) @lucifer4073

• [ENH] add a difference transformer to series transformations (#2729) @TinaJin0228

• [ENH] Starting Self Supervised Model with first example (#2385) @hadifawaz1999

• [ENH] Series transformer pipeline and datatype list tidy (#2830) @MatthewMiddlehurst

• [ENH,REF] Unequal length collection transforms (#2903) @MatthewMiddlehurst

Refactored

• [REF] Refactor smoothing transformers into a sub-package (#2783) @MatthewMiddlehurst

• [ENH,REF] Unequal length collection transforms (#2903) @MatthewMiddlehurst

Unit Testing

Bug Fixes

• [BUG] Adds test case for using csr_matrix with pickle - invalidates #2210 (#2843) @patrickzib

Documentation

• [DOC] Fix Broken [Source] Link and Improve Documentation for suppress_output() (#2677) @Kaustbh

Enhancements

• [ENH] Forecasting testing (#2891) @MatthewMiddlehurst

Maintenance

• [MNT] stop segmenters changing state in predict (#2526) @TonyBagnall

• [MNT] Drop Python 3.9 (#2845) @MatthewMiddlehurst

• [MNT] Fixing CI issues and separate doctests from regular CI runners (#2902) @MatthewMiddlehurst

Visualisations

Maintenance

• [MNT] stop segmenters changing state in predict (#2526) @TonyBagnall

Other

Documentation

• [DOC] Add implementation references (#2748) @MatthewMiddlehurst

• [DOC] Documentation improvement of BaseSeriesEstimator and BaseAeonEstimator (#2612) @kevinzb56

• [DOC,MNT] Add supporting devs to webpage and consolidate maintenance workflows (#2797) @MatthewMiddlehurst

• [DOC] Update the projects lists (#2146) @TonyBagnall

Maintenance

• [MNT] Use GPU installs for periodic tests (#2747) @MatthewMiddlehurst

• [MNT] Lower bound keras to discard old bugged versions of early keras3 (#2808) @hadifawaz1999

• [DOC,MNT] Add supporting devs to webpage and consolidate maintenance workflows (#2797) @MatthewMiddlehurst

• [MNT] Run PR tests when merging on any branch (#2809) @MatthewMiddlehurst

• [MNT] Update Slack link (#2877) @MatthewMiddlehurst

• [MNT] PR open bot comment changes (#2878) @MatthewMiddlehurst

• [MNT] codecov badge fix (#2908) @MatthewMiddlehurst

• [MNT] Fix codeowners for AD metrics (#2784) @SebastianSchmidl

Other

• [GOV] Supporting Developer role (#2775) @MatthewMiddlehurst

• [GOV] Remove inactive developers (#2776) @MatthewMiddlehurst

• [ENH] Correct broken source links for set_params and get_params in Aeon estimator docs (#2754) @Kaustbh

• [GOV] Use the NumFOCUS code of conduct and rename aeon workgroup (#2852) @MatthewMiddlehurst

Contributors

The following have contributed to this release through a collective 75 GitHub Pull Requests:

@aadya940, @adityagh006, @Ahmed-Zahran02, @AnaghDeshpande, @aryanpola, @baraline, @hadifawaz1999, @inclinedadarsh, @Kaustbh, @kavya-r30, @kevinzb56, @lucifer4073, @MatthewMiddlehurst, @nMaax, @patrickzib, @pattplatt, @Ramana-Raja, @saadaltohamy, @SebastianSchmidl, @shinymack, @SomtoOnyekwelu, @steenrotsman, @tanishy7777, @TinaJin0228, @TonyBagnall