Hands-On Python Deep Learning for the Web

Compiling and training a Keras neural network

We are now ready to compile and train the neural network. To compile the neural network, we use the following code:

model.compile(loss=keras.losses.categorical_crossentropy,              optimizer=keras.optimizers.Adam(),              metrics=['accuracy'])

In our model, which we compiled in the previous block of code, we have set categorical cross-entropy as the loss function; the optimizer function used is the Adam optimizer and the metric for evaluation is accuracy.

We then train the neural network with the fit() method of the Keras model object:

model.fit(x_train, y_train,          batch_size=100,          epochs=10,          verbose=2,          validation_split=0.2)

It is recommended that you perform a split of the training data into further validation ...