A neural network whereby the unsupervised learning algorithm is trained to set the output target values equal to the inputs. The autoencoder’s goals are to reduce the number of random variables under consideration, such that the input is represented in fewer dimensions (see Dimension Reduction).
