Using K-fold cross-validation in Keras on the data of my model

Question

I would like to use K-fold cross-validation on my data of my model.

My codes in Keras is :

a = np.array(  
[[283, 95, 72, 65],
[290, 100, 80, 72],
[120,170,130,122],
[100,230,110,200],
[300,100,200,500]]
)
X = a[:,0:2]
Y = a[:,3]

from sklearn.model_selection import KFold, cross_val_score
k_fold = KFold(n_splits=3)    
model = models.Sequential()
model.add(Dense(12, input_shape=(3,)))
model.add(LeakyReLU())
model.summary()

cross_val_score(model,X,Y)

But, It makes this error:

If no scoring is specified, the estimator passed should have a 'score' method. The estimator does not.

And when I select a scoring parameter as:

cross_val_score(model,X,Y, scoring= 'accuracy')

It makes another error:

TypeError: Cannot clone object '' (type ): it does not seem to be a scikit-learn estimator as it does not implement a 'get_params' methods.

How can I use K-fold cross-validation on this model?

Thank you

Answer 1

The cross_val_score seems to be dependent on the model being from sk-learn and having a get_params method. Since your Keras implementation does not have this, it can't provide the necessary information to do the cross_val_score. Try the manual k-fold cross validation found here: https://machinelearningmastery.com/evaluate-performance-deep-learning-models-keras/

Answer 2

There is a nice SKLearn wrapper for keras:

https://keras.io/scikit-learn-api/

I have not personally tested cross validation using the wrapper, but I anticipate it will work since everything else I've tried with the wrapper (no matter how strange) worked as expected.