Now that we defined encoder and decoder, we define the model which takes images as input and returns the output of the decoder which is the reconstructed image:

model = Model(inputs=input_image, outputs=decoder)

Let's look at a summary of the model:

model.summary()

________________________________________________________________
Layer (type)                Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 784)               0         
_________________________________________________________________
dense (Dense)                (None, 32)                25120     
_________________________________________________________________
dense_1 (Dense)              (None, 784)               25872     
=================================================================
Total params: 50,992
Trainable params: 50,992
Non-trainable params: 0
_________________________________________________________________

Compile the model with loss as a binary cross-entropy and minimize the loss using the adadelta optimizer:

model.compile(optimizer='adadelta', loss='binary_crossentropy')

Now let's train the model.

Generally, we train the model as model.fit(x,y) where x is the input and y is the label. But since autoencoders reconstruct their inputs, the input and output to the model should be the same. So, here, we train the model as model.fit(x_train, x_train):

model.fit(x_train, x_train, epochs=50, batch_size=256, shuffle=True, validation_data=(x_test, x_test))