Autoencoders for Representation Learning