[ENH] NearestCentroidClassifier#3527
Open
chrisholder wants to merge 2 commits into
Open
Conversation
aeon-actions-bot
Bot
added
classification
Contributor
Thank you for contributing to
|
Add a nearest-centroid (Rocchio) time series classifier. It builds one centroid per class by averaging the class's training series, then assigns a test series to the nearest class centroid under a chosen elastic distance. - Named `NearestCentroidClassifier` to avoid colliding with `sklearn.neighbors.NearestCentroid` (mirrors `KNeighborsTimeSeriesClassifier`). - Defaults to `distance="dtw"` with `average_method=None`, which resolves to DBA (petitjean) for ordinary distances and to the gradient-based soft barycentre for soft distances. A soft distance and `average_method="soft"` must be used together; mismatches raise `ValueError`. - Centroids are computed via `elastic_barycenter_average` (or `mean_average` for `"mean"`), so the soft-coupling validation is shared with the averaging module. - Full numpydoc docstring, tags, `_get_test_params`, unit tests, and an entry in the classification API reference. Passes the general estimator checks.
- add a `random_state` constructor parameter, forwarded to the barycentre
averaging (mirroring how `n_jobs` is passed). Previously a seed could only be
set via the undiscoverable `average_params={"random_state": ...}`. It affects
the stochastic averaging methods and random init only.
- add tests: multivariate fit+predict end-to-end, and reproducibility under a
fixed random_state with a stochastic averaging method.
chrisholder
force-pushed
the
soft-barycentre-averaging
branch
from
13fcd3c to
4a175f4
Compare
chrisholder
force-pushed
the
nearest-centroid-classifier
branch
from
2c2f901 to
82305d1
Compare
chrisholder
requested review from
MatthewMiddlehurst and
TonyBagnall
as code owners
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reference Issues/PRs
Fourth and final PR of the soft-distance stack. Stacked on #3526 (soft barycentre averaging) → #3508 → #3483. Please review/merge those first; this branch targets #3526 and the diff shrinks as they merge.
What does this implement/fix?
Adds
NearestCentroidClassifier, a nearest-centroid (Rocchio) time series classifier. It computes one centroid per class by averaging that class's training series, then classifies a new series by the nearest class centroid under a chosen elastic distance.NearestCentroidClassifierto avoid colliding withsklearn.neighbors.NearestCentroid(the same reason aeon's KNN isKNeighborsTimeSeriesClassifier).distance="dtw"andaverage_method=None, which resolves to DBA (petitjean) for ordinary distances and to the gradient-based soft barycentre for soft distances.average_method="soft"; explicitly pairing a soft distance with a non-soft averaging method (or vice versa) raisesValueError. Centroids are computed throughelastic_barycenter_average, so this validation is shared with the averaging module rather than re-implemented._tags,_get_test_params, a runnableExamplesdoctest, and an entry in the classification API reference.Does your contribution introduce a new dependency?
No.
Test plan
test_nearest_centroid.py: per-class centroid shapes, label prediction, the soft auto-promote, the validation matrix (soft+hard and hard+soft both raise), the DBA default, and that mean vs DBA centroids differ on warped data.check_estimator, 22/22) and the doctest.