CODEWITHSUNDEEP
scikit-learnneural-networkgrid-search
Dela
Dela

Reputation: 115

Neural network grid search value error

I am trying to run a grid search for a neural network but I keep getting some weird errors. my algorithm looks like:

parameters={'learning_rate':["constant", "invscaling", "adaptive"], 
                 'hidden_layer_sizes': (156,), 'alpha': [10.0 ** -np.arange(1, 7)], 
                 'activation': ["logistic", "relu", "Tanh"]}
grid= GridSearchCV(MLPClassifier(),parameters, n_jobs=-1, cv=10)
grid.fit(train_x, train_y)

The error message I get is:

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

I have tried to also use just 1 value wit regards to the activation and learning_rate but the problem seem to persist. Is there anything I am not doing well, please?

Upvotes: 1

Views: 116

Answers (1)

seralouk
seralouk

Reputation: 33147

I spotted 2 mistakes in your code.

First: The alpha parameters should be contained in a pure list. Using List Comprehension, the answer is as follows.

Second: In the 'activation': ["logistic", "relu", "Tanh"]} the Tanh should be replaced with tanh.

The following code should work fine:

Replace: 

'alpha': [10.0 ** -np.arange(1, 7)]
'activation': ["logistic", "relu", "Tanh"]

With:

'alpha': [10.0 ** -i for i in range(1,7)]
'activation': ["logistic", "relu", "tanh"]

Putting everything together:

parameters={'learning_rate':["constant", "invscaling", "adaptive"], 
             'hidden_layer_sizes': (156,), 'alpha': [10.0 ** -i for i in range(1,7)], 
             'activation': ["logistic", "relu", "tanh"]}

grid= GridSearchCV(MLPClassifier(), parameters, n_jobs=-1, cv=10)

grid.fit(train_x, train_y)

Upvotes: 1

Related Questions