Scikit-Learn GridSearchCV failing on on a gensim LDA model

Question

This is the code for creating the model :

import gensim
NUM_TOPICS = 4
ldamodel = gensim.models.ldamodel.LdaModel(corpus,num_topics = 
NUM_TOPICS,id2word=dictionary,passes=100)
ldamodel.save('model5.gensim')
topics = ldamodel.print_topics(num_words=4)
print(topics)

This is the code for GridSearchCV :

search_params = {'n_components': [4, 6, 8, 10, 20], 'learning_decay': [.5, .7, .9]}


# Init Grid Search Class
model = GridSearchCV(ldamodel, param_grid=search_params)

# Do the Grid Search
model.fit(data_vectorized)

This is the output :

*---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-108-1a35c49ac19e> in <module>
      9 
     10 # Do the Grid Search
---> 11 model.fit(data_vectorized)
~\AppData\Local\Continuum\anaconda3\lib\site-packages\sklearn\model_selection\_search.py in fit(self, X, y, groups, **fit_params)
    627 
    628         scorers, self.multimetric_ = _check_multimetric_scoring(
--> 629             self.estimator, scoring=self.scoring)
    630 
    631         if self.multimetric_:
~\AppData\Local\Continuum\anaconda3\lib\site-packages\sklearn\metrics\_scorer.py in _check_multimetric_scoring(estimator, scoring)
    471     if callable(scoring) or scoring is None or isinstance(scoring,
    472                                                           str):
--> 473         scorers = {"score": check_scoring(estimator, scoring=scoring)}
    474         return scorers, False
    475     else:
~\AppData\Local\Continuum\anaconda3\lib\site-packages\sklearn\metrics\_scorer.py in check_scoring(estimator, scoring, allow_none)
    399     if not hasattr(estimator, 'fit'):
    400         raise TypeError("estimator should be an estimator implementing "
--> 401                         "'fit' method, %r was passed" % estimator)
    402     if isinstance(scoring, str):
    403         return get_scorer(scoring)
TypeError: estimator should be an estimator implementing 'fit' method, <gensim.models.ldamodel.LdaModel object at 0x000002121E55D3C8> was passed*

Answer 1

You are trying to use GridSearchCV object from a scikit-learn package which requires the model object on which it is run to implement certain methods (as in the error message: fit method in particular). Since scikit-learn is not related in any way to gensim you need to ensure that they are compatible by subclassing an Estimator class in scikit-learn and encapsulating gensim training in the fit method.

Also, it does not seem to me in the LdaModel documentation that it uses the parameters ( n_components , learning_decay ) that you are attempting to search for. You can only search for the values of the parameters that the model uses.