Hands-On Artificial Intelligence for Beginners

Boltzmann machines

Boltzmann machines are a general class of models that contain take binary vectors as input and units that assign a probability distribution to each of those binary vectors. As you can see in the following diagram, each unit is dependent on every other unit:

A Boltzmann machine uses something called an energy function, which is similar to a loss function. For any given vector, the probability of a particular state is proportional to each of the energy function values. To convert this to an actual probability distribution, it's necessary to renormalize the distribution, but this problem becomes another intractable problem. ...