A Boltzmann machine is a type of random neural network and recurrent neural network invented by Geoffrey Hinton and Terry Sejnowski in 1985.

The Boltzmann machine can be viewed as a random process that can generate the corresponding Hopfield neural network. It is the earliest neural network that can learn internal expressions and can express and solve complex combinatorial optimization problems.

Due to the locality and Hebbian nature of the training algorithms, as well as their parallelism similar to simple physical processes, the learning method will be efficient enough to solve practical problems if the connection method is constrained. It is named after the Boltzmann distribution, which is used for the sampling function of the Boltzmann machine.