Inference with Primula

• Exact inference by model compilation into Bayesian Networks, and use of Bayesian Network inference engines, notably the SamIam system.
• Exact inference by model compilation into Arithmetic Circuits using the ACE system.
• Approximate inference using importance sampling

The Inference Module

is used to enter evidence, specify queries, call the built-in sampling inference, as well as the integrated inference with ACE system.

Alternatively, the domain specific network can be exported as a Bayesian Network model, and inference is performed in an external Bayesian Network system. In particular, Primula is directly interfaced with SamIam:

• Learning
• Interpreting Markov Logic Networks