[BUG] ShapeletTransformPyts().fit(x) throws "ValueError: ShapeletTransformPyts requires `y` in `fit`."
navaldroid opened this issue · comments
Describe the bug
For some reason, ShapeletTransformPyts (and also ShapeletTransform) requires the y parameter in their fit() method, despite the documentations for fit(X, y=None) reports it as optional
import numpy as np
from sktime.transformations.panel.shapelet_transform import ShapeletTransformPyts
x = np.random.random(10000)
model = ShapeletTransformPyts()
fitted = model.fit(x)
>>> ValueError: ShapeletTransformPyts requires `y` in `fit`.Versions
Python 3.10.10
numpy==1.26.4
pyts==0.13.0
sktime==0.29.0
scikit-base==0.7.7
This is not a bug, but possibly a documentation improvement point.
ShapeletTransformPyts indeed requires y in fit. This is as in the interfaced transformer, from pyts, see https://pyts.readthedocs.io/en/stable/generated/pyts.transformation.ShapeletTransform.html
The documentation could be clearer - some transformations do require y in fit, some do not. The current docstring says "some transformers require this, see class docstring for details".
Do you have a suggestion how this could be improved? Perhaps pointing to the requires_y tag?
@navaldroid, thanks for your help - although I do not think it is as easy - or unproblematic - to remove the default.
The default needs to be there in fit, because some transformations do not require y.
The docstring also applies to all transformations at once, at the moment - so a change there would make it false for the remaining transformer.s