The Boltzmann machine is an early form of neural network based on spin glass models in physics, which also have parallels in human brain mechanisms. Boltzman machines are often used as an autoencoder. Boltzman machines have two types of nodes (or layers): visible (which can be used for input and output) and hidden (which often act as low-demiensional encodings of inputs). In a standard Boltzmann machine all nodes are fully interconnected, but in Restricted Boltzmann Machine there are only connections between the layers.
Defined on page 119
Used on pages 109, 119, 120, 123, 277, iii
Boltzmann Machine