Add support for returning only returning x in PipelineElement's transform api

[Paul-B98]

I like to propose a bahivour where only not falsy elements are returned. The reason for this is that some transformers like the ColumnTransformer break when they get more elements than x, as in the case of the PipelineElement. This would also be more compatible with Sklearn's preprocessing transformers, which in most cases also only run X.

photonai/photonai/base/photon_elements.py

Lines 528 to 557 in ea31ea7

	def transform(self, X: np.ndarray, y: np.ndarray = None, **kwargs) -> (np.ndarray, np.ndarray, dict):
	"""
	Calls transform on the base element.
	In case there is no transform method, calls predict.
	This is used if we are using an estimator as a preprocessing step.
	Parameters:
	X:
	The array-like data with shape=[N, D], where N is the
	number of samples and D is the number of features.
	y:
	The truth array-like values with shape=[N], where N is
	the number of samples.
	**kwargs:
	Keyword arguments, passed to base_element.transform.
	Returns:
	(X, y) in transformed version and original kwargs.
	"""
	if self.batch_size == 0:
	Xt, yt, kwargs = self.__transform(X, y, **kwargs)
	else:
	Xt, yt, kwargs = self.__batch_transform(X, y, **kwargs)
	if all(hasattr(data, "shape") for data in [X, Xt]) and all(len(data.shape) > 1 for data in [X, Xt]):
	self.reduce_dimension = (Xt.shape[1] < X.shape[1])
	return Xt, yt, kwargs

[RLeenings]

Dear Paul-B98,
we have extended the transformer interface because, other than scikit-learn, we have an extended definition of the pipeline:

1.) we allow to make changes to the dataset as a whole, i.e. we allow to adapt the amount of data (i.e. through sample pairing) or to relabel items during the process (so that changes to the y are needed)
2.) we also allow to stream and edit additional data (that is not part of the feature set) through the pipeline. This additional information about the samples might be of use to a specific algorithm in the pipeline. For example when you need to have access to cluster assignments or meta information of medical data or similar data, although not every algorithm in the pipeline (particular those standard algorithms such as scaling, dimensionality reduction, etc.) need this information. The additional data is part of the nested cross-validation data split, so that the additional data is this is handled for the user.

I hope that makes sense to you.
Ramona