'FastICA' object has no attribute 'predict'

[skp1996]

When I do hyperparameter tuning of FastICA using GridSearchCV, It gives the error 'FastICA' object has no attribute 'predict'
The code is been executed on Google Colab.

from sklearn.datasets import load_digits
from sklearn.decomposition import FastICA
from sklearn.model_selection import GridSearchCV

X, y = load_digits(return_X_y=True)
ica=FastICA()
param_grid = {'n_components':[2,3,4,5,6,7,8,9,10,11,12,13]}
gs=GridSearchCV(ica,param_grid,cv=10,scoring='accuracy')
gs.fit(X,y)


AttributeError                            Traceback (most recent call last)
<ipython-input-8-1120d0fa4757> in <module>()
      5 param_grid = {'n_components':[2,3,4,5,6,7,8,9,10,11,12,13]}
      6 gs=GridSearchCV(ica,param_grid,cv=10,scoring='accuracy')
----> 7 gs.fit(X,y)

14 frames
/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_scorer.py in _cached_call(cache, estimator, method, *args, **kwargs)
     50     """Call estimator with method and args and kwargs."""
     51     if cache is None:
---> 52         return getattr(estimator, method)(*args, **kwargs)
     53 
     54     try:

AttributeError: 'FastICA' object has no attribute 'predict'

[NicolasHug]

FastICA isn't a predictor: it doesn't predict a target y, so it doesn't have a predict() method.
It's just a transformer, meaning it can only transform X via a transform() method

[MikeB2019x]

Is there any way using GridSearchCV to make the target the original input? I am facing this problem with KernelPCA ie. the measure of how well it 'scores' is whether the inverse transform has a low mse compared to the input. I can't figure out how to do this creating my own scorer b/c I have to pass in the original input and the transformed input. Per the docs for making your own scoring function it would sketch out like this:

1. estimator: KernelPCA()
2. X: kpca.fit(data)
3. y: data
   and my scorer function would return: neg_mean_squared(y - X)