The key idea of restricted Boltzmann machine:
A restricted Boltzmann machine which consists of a layer of stochastic binary “visible” units that represent binary input data connected to a layer of chochastic binary hidden units that learn to model significant nonindependencies between the visible units.
There are undirected connections between visible and hidden units but no visible-visible or hidden-hidden connections.
An RBM is type of Markov random field (MRF) but differs from most of MRFs in several ways: it has a bipartite connectivity graph, it does not usually share weights between different units, and a subset of the variables are unobserved, even during training.