Autoencoder is a network that learns to encode itself. The neural network is trained to copy its input at the output layer. In this code, the two autoencoder layers are stacked together and the overall architecture is build to perform classification.