[DOC] Fitting MultiRocketHydraClassifier with class_weight="balanced_subsample" gives an error

[dschrempf]

Describe the bug

Fitting MultiRocketHydraClassifier with class_weight="balanced_subsample" gives an error.

Steps/Code to reproduce the bug

from aeon.classification.convolution_based import MultiRocketHydraClassifier
from aeon.testing.data_generation import make_example_3d_numpy

X, y = make_example_3d_numpy(n_cases=10, n_channels=1, n_timepoints=12, random_state=0)
clf = MultiRocketHydraClassifier(random_state=0, class_weight="balanced_subsample")
clf.fit(X, y)

Expected results

Classifier is fitted.

Actual results

Traceback (most recent call last):
  File "/home/mext/mext-dom-aeon/spikes/eeg/test.py", line 6, in <module>
    clf.fit(X, y)
  File "/home/mext/mext-dom-aeon/.venv/lib/python3.9/site-packages/aeon/classification/base.py", line 114, in fit
    self._fit(X, y)
  File "/home/mext/mext-dom-aeon/.venv/lib/python3.9/site-packages/aeon/classification/convolution_based/_mr_hydra.py", line 127, in _fit
    self.classifier.fit(Xt, y)
  File "/home/mext/mext-dom-aeon/.venv/lib/python3.9/site-packages/sklearn/base.py", line 1466, in wrapper
    estimator._validate_params()
  File "/home/mext/mext-dom-aeon/.venv/lib/python3.9/site-packages/sklearn/base.py", line 666, in _validate_params
    validate_parameter_constraints(
  File "/home/mext/mext-dom-aeon/.venv/lib/python3.9/site-packages/sklearn/utils/_param_validation.py", line 95, in validate_parameter_constraints
    raise InvalidParameterError(
sklearn.utils._param_validation.InvalidParameterError: The 'class_weight' parameter of RidgeClassifierCV must be an instance of 'dict', a str among {'balanced'} or None. Got 'balanced_subsample' instead.

Versions

System: python: 3.9.20 (main, Oct 16 2024, 04:36:33) [Clang 18.1.8 ] executable: /home/.../.venv/bin/python3 machine: Linux-6.8.0-49-generic-x86_64-with-glibc2.39 Python dependencies: aeon: 1.0.0 pip: None setuptools: 75.6.0 scikit-learn: 1.5.2 numpy: 1.26.4 numba: 0.60.0 scipy: 1.13.1 pandas: 2.2.3

[MatthewMiddlehurst]

Thanks for the report, I will try take a look.

[baraline]

I think the issue is that balanced subsample is a strategy only supported by ensembles if I remember correctly?

The current classifier used here is a ridgeCV which can't support this strategy. I'm not 100% familiar with Hydra MR, but either :

• the aeon documentation is wrong and balanced subsample is not supported.
• the classification strategy is not correctly implemented for Hydra MR (since don't remember whether the algorithm should be an ensemble classifier or not)

[MatthewMiddlehurst]

This is passed directly to the classifier, seems like a documentation issue.

[Akhil-Jasson]

I reviewed the documentation, but I’m unable to find any reference to balanced_subsample being supported by MultiRocketHydraClassifier.

Could you point me to the specific part of the documentation where this is mentioned?

[baraline]

The issue was that balanced_subsample is not supported by the default RidgeCV estimator used in mr hydra. So the goal of the issue is to change the docstring here to describe that the class_weight parameter is given to the sklearn estimator and the supported values will be the ones of the estimators.

We could also add a check for that, but that might be one of the goal of #1777.

[Akhil-Jasson]

Wouldn't this particular issue be better categorized as a DOC rather than a BUG?

[dschrempf]

In retrospect, I would say yes. I am changing the title.

[Akhil-Jasson]

Hi @baraline, could you please take a look at my PR when you get a chance? Let me know if there's anything else I should address or look into. Thanks!

[MatthewMiddlehurst]

I think this case has been fixed. Thanks, will make sure have new PRs check for this before merging.