Registry / ai-ml / aeon
library1.4.0pypypiunverified

Aeon is an open-source toolkit for time series machine learning, offering transformers, classifiers, regressors, clusterers, and more. It is designed to be compatible with scikit-learn and provides a unified API for time series tasks. The current version is 1.4.0, with recent releases every few months.

pip install aeon
INSTALL
IMPORT
SIG · AEON
A
aeon
ai-mlpythonv1.4.0
harness data pending
Install & Compatibility
Where this runs

No compatibility data collected yet for this library.

Code
Verified usage

Verified import paths — ran on the pinned version, not inferred.

TimeSeriesForestClassifier
from aeon.classification.forest import TimeSeriesForestClassifier
from aeon.classification.forest import TimeSeriesForestClassifier

Load the UnitTest dataset, train a ROCKET classifier, and evaluate accuracy.

import numpy as np from aeon.classification.convolution_based import ROCKETClassifier from aeon.datasets import load_unit_test X_train, y_train = load_unit_test(split='train') X_test, y_test = load_unit_test(split='test') classifier = ROCKETClassifier(num_kernels=1000, random_state=42) classifier.fit(X_train, y_train) y_pred = classifier.predict(X_test) accuracy = np.mean(y_pred == y_test) print(f'Accuracy: {accuracy:.3f}')
Debug
Known issues
breakingaeon 1.0.0 introduced significant API changes, including removal of deprecated classes and renaming of modules (e.g., 'classifiers' -> 'classification'). Code written for earlier versions may break.
fix
Update imports to use the new module structure; refer to migration guide in aeon documentation.
affects: <1.0.0 to 1.0.0+
deprecatedThe 'sktime' compatibility layer is deprecated since aeon 1.1.0 and will be removed in a future release.
fix
Use native aeon imports instead of sktime wrappers.
affects: >=1.1.0
gotchaaeon expects 3D numpy arrays of shape (n_instances, n_channels, n_timepoints) for collections. If you pass 2D arrays, the behavior may be inconsistent or raise errors.
fix
Reshape your data to 3D (e.g., X = X[np.newaxis, :, :] for single channel).
affects: all
gotchaMany estimators have a random_state parameter but do not guarantee full reproducibility across different hardware or library versions due to underlying NumPy random number generation.
fix
Set random_state and use deterministic algorithms where possible.
affects: all
Upgrade
Version history
1.4.0latest on PyPI · released Mar 24, 2026
Audit
Dependencies
scikit-learnrequiredCore dependency for base estimators and compatibility
numpyrequiredArray operations
pandasrequiredDataFrame handling
Agent activity
64 hits · last 30 days
node
56
OpenAI (training)
1
Resources
aeon — pip install aeon · libregistry