Presentation
GMtoolbox currently allows to:
- represent a graphical model as a factor graph, such that:
- infer marginals using evidence (observation).
The inference relies on the Generalized Belief Propagation algorithm parent-to-child (see JS Yedidia, WT Freeman, Y Weiss, “Constructing Free-Energy Approximations and Generalized Belief Propagation Approximations,” IEEE Transactions on Information Theory (2005) vol. 51, pp. 2282-2312) [PDF].