[ENH] Import RandomChannelSelector through init (#1933)

* remove Theta

* docstring

* docstring

* fix bug

* all in channel selection

* docstring

Author: TonyBagnall

aeon/transformations/collection/channel_selection/__init__.py

@@ -5,7 +5,12 @@
 collection using transform.
 """
 
-__all__ = ["ChannelScorer", "ElbowClassPairwise", "ElbowClassSum"]
+__all__ = [
+    "ChannelScorer",
+    "ElbowClassPairwise",
+    "ElbowClassSum",
+    "RandomChannelSelector",
+]
 
 
 from aeon.transformations.collection.channel_selection._channel_scorer import (
@@ -15,3 +20,6 @@
     ElbowClassPairwise,
     ElbowClassSum,
 )
+from aeon.transformations.collection.channel_selection._random import (
+    RandomChannelSelector,
+)

aeon/transformations/collection/channel_selection/_channel_scorer.py

@@ -7,6 +7,9 @@
 from aeon.classification.convolution_based._rocket_classifier import RocketClassifier
 from aeon.transformations.collection.channel_selection.base import BaseChannelSelector
 
+__maintainer__ = ["TonyBagnall"]
+__all__ = ["ChannelScorer"]
+
 
 class ChannelScorer(BaseChannelSelector):
     """Channel scorer performs channel selection using a single channel classifier.

aeon/transformations/collection/channel_selection/_random.py

@@ -7,6 +7,7 @@
 from aeon.transformations.collection.channel_selection.base import BaseChannelSelector
 
 __maintainer__ = ["TonyBagnall"]
+__all__ = ["RandomChannelSelector"]
 
 
 class RandomChannelSelector(BaseChannelSelector):
@@ -15,13 +16,23 @@ class RandomChannelSelector(BaseChannelSelector):
     Parameters
     ----------
     p: float, default 0.4
-        proportion of channels to keep. If p*len(X) is non integer it is rounded up
+        proportion of channels to keep. If p*n_channels is non integer it is rounded up
         to the nearest integer.
 
     Attributes
     ----------
     channels_selected_ : list[int]
         List of channels selected in fit.
+
+    Examples
+    --------
+    >>> import numpy as np
+    >>> from aeon.transformations.collection.channel_selection import RandomChannelSelector # noqa
+    >>> X = np.random.rand(10, 10, 100)
+    >>> selector = RandomChannelSelector(p=0.4)
+    >>> XNew = selector.fit_transform(X)
+    >>> XNew.shape
+    (10, 4, 100)
     """
 
     _tags = {
@@ -41,7 +52,7 @@ def __init__(self, p=0.4, random_state=None):
     def _fit(self, X, y):
         """Randomly select channels to retain."""
         rng = check_random_state(self.random_state)
-        to_select = math.ceil(self.p * len(X))
+        to_select = math.ceil(self.p * X.shape[1])
         self.channels_selected_ = rng.choice(
             list(range(X[0].shape[0])), size=to_select, replace=False
         )