Now let's build the same autoencoder in Keras.
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()
- First, we import the keras libraries and define hyperparameters and layers:
import kerasfrom keras.layers import Densefrom keras.models import Sequentiallearning_rate = 0.001n_epochs = 20batch_size = 100n_batches = int(mnist.train.num_examples/batch_sizee# number of pixels in the MNIST image as number of inputsn_inputs = 784n_outputs = n_i# number of hidden layersn_layers = 2# neurons in each hidden layern_neurons = [512,256]# add decoder ...
