Gibbs sampling

       Gibbs sampling is an approximation technique for computing the conditional probability through a stochastic simulation, and it is regarded as one of the inference methods (Metropolis et al., 1953; Gelfand & Smith, 1990; Neal, 1992; York, 1992). Gibbs sampling accumulates a sample by visiting nodes in turn and changing the output value of the node. Each visit corresponds to a new state. The value of each node is updated according to the probability of the node conditioned by the values of all the other nodes, and the convergence of the approximation is guaranteed if, in a given visiting scheme, each node can be visited infinitely often.