Continuous Distributions
For the moment only Gaussian Distributions are available and can only be used with the MCMC Inference Engine. Plans are to develop full Gaussian support for any inference engine (and learning engine) as well as mixture of gaussians which will allow to approximate any distribution by using a sum of gaussians. (Developper : Kosta Gaitanis)
To achieve this, Gaussian Potentials must be implemented :
- CG Potential : for pure gaussian BNs
- SG Potentials : for stable hybrid BNs
This feature is under development since some time and will be supposedly finished during the year 2007.