Visualizing Autoencoders with Tensorflow.js

An autoencoder is a type of neural network that is comprised of two functions: an encoder that projects data from high to low dimensionality, and a decoder that projects data from low to high dimensionality. To understand how these two functions work, let's consider the following images: While the encoder reduces the dimensionality of input data, the decoder projects samples from low dimensionality back to higher dimensionality. For example, if one constructs a decoder that projects data from 2 dimensions to 748 dimensions, it becomes possible to project arbitrary positions in a two dimensional plane into a 748 pixel image. Click around in the figure below to see how a decoder projects from 2 to 748 dimensions. An autoencoder is a neural network that combines the encoder and decoder discussed above into a single model that projects input data to a lower-dimensional embedding (the encode step), and then projects that lower-dimensional data back to a high dimensional embedding (the decode step).