Stacked autoencoder in Keras

Now let's build the same autoencoder in Keras.

Note

We clear the graph in the notebook using the following commands so that we can build a fresh graph that does not carry over any of the memory from the previous session or graph:

 

tf.reset_default_graph()keras.backend.clear_session()

1. First, we import the keras libraries and define hyperparameters and layers:

import keras
from keras.layers import Dense
from keras.models import Sequential

learning_rate = 0.001
n_epochs = 20
batch_size = 100
n_batches = int(mnist.train.num_examples/batch_sizee
# number of pixels in the MNIST image as number of inputs
n_inputs = 784
n_outputs = n_i
# number of hidden layers
n_layers = 2
# neurons in each hidden layer
n_neurons = [512,256]
# add decoder layers:
n_neurons.extend(list(reversed(n_neurons)))
n_layers = n_layers * 2

2. Next, we build a sequential model and add dense layers to it. For a change, we use relu activation for the hidden layers and linear activation for the final layer...