Mastering TensorFlow 1.x

Functional API to add layers to the Keras Model

In the functional API, the layers are created first in a functional manner, and then while creating the model, the input and output layers are provided as tensor arguments, as we covered in the previous section. 

Here is an example:

1. First, create the input layer:

input = Input(shape=(64,))

2. Next, create the dense layer from the input layer in a functional way:

hidden = Dense(10)(inputs)

3. In the same way, create further hidden layers building from the previous layers in a functional way:

hidden = Activation('tanh')(hidden)hidden = Dense(10)(hidden)output = Activation('tanh')(hidden)

4. Finally, instantiate the model object with the input and output layers:

model = Model(inputs=input, outputs=output) ...