In the functional model, we must create and define an input layer, which specifies the shape of the input data. The input layer takes a shape
argument that is a tuple, which indicates the dimensionality of the input data. When the input data is one-dimensional (for example, for a multilayer perceptron), the shape must leave space for the shape of the mini-batch size, which is determined while splitting the data when training the network. The shape tuple is always defined with an open last dimension when the input is a one-dimensional example (32).
In the following code, we define the first layer:
from keras.layers import Input visible = Input(shape=(32,))
We connect the layers together:
visible = Input(shape=(32,)) hidden = Dense(32)(visible)