ABSTRACT
An autoencoder is a neural network that is trained to learn efficient representations of the input data. Since autoencoders are, fundamentally, feedforward deep learning models, they come with all the benefits and flexibility that deep learning models provide. An autoencoder has a structure very similar to a feedforward neural network; however, the primary difference when using in an unsupervised context is that the number of neurons in the output layer is equal to the number of inputs. When the autoencoder uses only linear activation functions and the loss function is mean squared error, then it can be shown that the autoencoder reduces to Principal components analysis. Sparse autoencoders are used to pull out the most influential feature representations. Training a denoising autoencoder is nearly the same process as training a regular autoencoder.
