[MNT] Wrong type annotations for aeon classes

[SebastianSchmidl]

Describe the issue

We started to add Python type hints to our estimators. However, some existing hints for (Aeon) classes are incorrect, e.g.

aeon/aeon/classification/sklearn/_rotation_forest_classifier.py

Lines 108 to 110 in be4af65

	remove_proportion: float = 0.5,
	base_estimator: Optional[Type[BaseEstimator]] = None,
	pca_solver: str = "auto",

• Type[BaseEstimator] references the type of the class (thus, estimatorFactory = BaseEstimator is of type Type[BaseEstimator])
• BaseEstimator references an instance of this class (thus, estimator = BaseEstimator() is of type BaseEstimator)
• more details about this destinction: https://mypy.readthedocs.io/en/stable/kinds_of_types.html#the-type-of-class-objects

Suggest a potential alternative/fix

Because the RotationForestClassifier works with an instance as base_estimator the correct type would be base_estimator: BaseEstimator

We need to check the code base for further such issues.

Additional context

No response

[SebastianSchmidl]

• aeon/aeon/regression/sklearn/_rotation_forest_regressor.py

  Line 107 in be4af65

  base_estimator: Optional[Type[BaseEstimator]] = None,

• check_random_state() also requires an instance of RandomState and not a class,
  e.g.

  aeon/aeon/classification/distance_based/_proximity_forest.py

  Line 92 in be4af65

  random_state: Union[int, Type[np.random.RandomState], None] = None,

  (and other locations)

[aryan0931]

i would like to work on this issue

[SebastianSchmidl]

Sure!

@aeon-actions-bot assign @aryan0931

[SebastianSchmidl]

@aryan0931 Are you still working on this?

[aryan0931]

no

[shinymack]

I would like to work on this issue.
@aeon-actions-bot assign @shinymack

[shinymack]

I was not able to find more of these incorrect type hints. The ones mentioned earlier are already fixed. Please let me know if there’s anything else you’d like me to check.

[SebastianSchmidl]

I can confirm that the mentioned issues have been resolved. However, I could find the following additional locations (not as important because private methods, but we are already on it, so why not fix it):

aeon/aeon/classification/sklearn/_continuous_interval_tree.py

Line 486 in 00235ac

def _find_splits_gain(self, node: type[_TreeNode], splits: list, gains: list):

aeon/aeon/regression/sklearn/_rotation_forest_regressor.py

Line 355 in 00235ac

def _predict_for_estimator(self, X, clf: int, pcas: type[PCA], groups):

aeon/aeon/classification/sklearn/_rotation_forest_classifier.py

Line 456 in 00235ac

def _predict_proba_for_estimator(self, X, clf: int, pcas: type[PCA], groups):

After those are fixed, we can close this IMO.

[SebastianSchmidl]

@shinymack thank you for your contributions!