Hands-On Machine Learning with C++

Autoencoders

Autoencoders represent a particular class of neural networks that are configured so that the output of the autoencoder is as close as possible to the input signal. In its most straightforward representation, the autoencoder can be modeled as a multilayer perceptron in which the number of neurons in the output layer is equal to the number of inputs. The following diagram shows that by choosing an intermediate hidden layer of a smaller dimension, we compress the source data into the lower dimension. Usually, values from this intermediate layer are a result of an autoencoder:

Now that we have learned about the linear and non-linear ...