The get_params for each estimator (anything that extends from BaseEstimator) will need to be overridden to ensure that all the correct parameters are displayed. This issue arises as the sklearn get_params implementation inspects the init of the current class naively. This inspection leads to two problems:
anything not in the __init__ will not be displayed
anything passed to a parent via **kwargs will not be shown.
We must handle these cases separately.
Correct parameters to display.
parameters in the __init__ method of the current class.
parameters passed to the __init__ method of the parent(s) class(es).
Note: the implementation of this will need to define where the true base parent class is. Currently it is just the Preparer level but we may traverse back more if changes are introduced to the parent class hierarchy.
attributes that are not in the __init__ signature but defined in the __init__ method (both current and parent).
Description
The get_params for each estimator (anything that extends from BaseEstimator) will need to be overridden to ensure that all the correct parameters are displayed. This issue arises as the sklearn get_params implementation inspects the init of the current class naively. This inspection leads to two problems:
Correct parameters to display.
Unit tests may take up most of the time.
Estimate: 2 days