Autoencoders

Abstract

Autoencoder is an unsupervised learning algorithm that uses neural networks to reconstruct the features of a dataset. Just like the unsupervised algorithms that we earlier discussed in the chapter on machine learning, autoencoders can be used to reduce the dimensionality of a dataset and to extract relevant features. Moreso, peculiar to autoencoders is the ability to generate more examples of the dataset after learning an internal representation (also called coding) that reconstructs the features of the inputs to the neural network.