Catalogue of Artificial Intelligence Techniques
Aliases: Gibbs Sampling
Keywords: Boltzmann machines
Categories: Problem Solving
Author(s): Judea Pearl
A method of computing probabilities by counting events in a series of simulation runs. In each run the occurrence of an event is determined by sampling a random distribution, reflecting the state of related events as well as the strength of interaction among them. In Boltzmann machines stochastic simulation is used together with Simulated Annealing to find the state of lowest energy.
- Pearl, J., Probabilistic Reasoning in Intelligent Systems:
networks of plausible inference
, Morgan Kaufmann, San Mateo, California, 1988.