Representation Learning With Autoencoder