Autoencoders

An autoencoder may be considered as a network that implements the identity function. It takes a data sample as the input and also considers the same sample as the output, thus behaving as a self-supervised model. Though the idea of a network aiming to replicate its input seems pointless, it may be used for many purposes such as