How to pass a parameter to Scikit-Learn Keras model function

Question

I have the following code, using Keras Scikit-Learn Wrapper, which work fine:

from keras.models import Sequential
from keras.layers import Dense
from sklearn import datasets
from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.model_selection import StratifiedKFold
from sklearn.model_selection import cross_val_score
import numpy as np


def create_model():
    # create model
    model = Sequential()
    model.add(Dense(12, input_dim=4, init='uniform', activation='relu'))
    model.add(Dense(6, init='uniform', activation='relu'))
    model.add(Dense(1, init='uniform', activation='sigmoid'))
    # Compile model
    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    return model


def main():
    """
    Description of main
    """


    iris = datasets.load_iris()
    X, y = iris.data, iris.target

    NOF_ROW, NOF_COL =  X.shape

    # evaluate using 10-fold cross validation
    seed = 7
    np.random.seed(seed)
    model = KerasClassifier(build_fn=create_model, nb_epoch=150, batch_size=10, verbose=0)
    kfold = StratifiedKFold(n_splits=10, shuffle=True, random_state=seed)
    results = cross_val_score(model, X, y, cv=kfold)

    print(results.mean())
    # 0.666666666667


if __name__ == '__main__':
    main()

The pima-indians-diabetes.data can be downloaded here.

Now what I want to do is to pass a value NOF_COL into a parameter of create_model() function the following way

model = KerasClassifier(build_fn=create_model(input_dim=NOF_COL), nb_epoch=150, batch_size=10, verbose=0)

With the create_model() function that looks like this:

def create_model(input_dim=None):
    # create model
    model = Sequential()
    model.add(Dense(12, input_dim=input_dim, init='uniform', activation='relu'))
    model.add(Dense(6, init='uniform', activation='relu'))
    model.add(Dense(1, init='uniform', activation='sigmoid'))
    # Compile model
    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    return model

But it fails giving this error:

TypeError: __call__() takes at least 2 arguments (1 given)

What's the right way to do it?

Answer 1

You can add an input_dim keyarg to the KerasClassifier constructor:

model = KerasClassifier(build_fn=create_model, input_dim=5, nb_epoch=150, batch_size=10, verbose=0)

Answer 2

Last answer does not work anymore.

An alternative is to return a function from create_model, as KerasClassifier build_fn expects a function:

def create_model(input_dim=None):
    def model():
        # create model
        nn = Sequential()
        nn.add(Dense(12, input_dim=input_dim, init='uniform', activation='relu'))
        nn.add(Dense(6, init='uniform', activation='relu'))
        nn.add(Dense(1, init='uniform', activation='sigmoid'))
        # Compile model
        nn.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
        return nn

    return model

Or even better, according to documentation

sk_params takes both model parameters and fitting parameters. Legal model parameters are the arguments of build_fn. Note that like all other estimators in scikit-learn, build_fn should provide default values for its arguments, so that you could create the estimator without passing any values to sk_params

So you can define your function like this:

def create_model(number_of_features=10): # 10 is the *default value*
    # create model
    nn = Sequential()
    nn.add(Dense(12, input_dim=number_of_features, init='uniform', activation='relu'))
    nn.add(Dense(6, init='uniform', activation='relu'))
    nn.add(Dense(1, init='uniform', activation='sigmoid'))
    # Compile model
    nn.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    return nn

And create a wrapper:

KerasClassifier(build_fn=create_model, number_of_features=20, epochs=25, batch_size=1000, ...)