Machine Learning Algorithms - Second Edition

Interfacing Keras to scikit-learn

Even if Keras provides many support functions, in some cases, it's easier to perform specific operations with scikit-learn. For example, we could be interested in an automated cross-validation grid search; as we're going to discuss in Chapter 17, Creating a Machine Learning Architecture, it's helpful to create a complete processing pipeline that performs all the preliminary operations on the dataset and then train the model. To solve these problems, Keras provides two wrappers, KerasClassifier and KerasRegressor, that implement the standard scikit-learn interfaces (for classifiers and regressors) but work with Keras models.

To show how to use these wrappers, let's perform a grid search on the model previously ...