A coin, with Multinomial distribution over outcomes, which are Flips.
A GibbsSampler that can also collapse some Parameters.
The expectation-maximization method of inference.
The outcome of a coin flip, with boolean value.
Beta distribution.
Alternative style of Dirichlet parameterized by 2 parents (mean,precision) rather than 1 (masses).
Create a new DirectedFactor, make it the "parent" generating factor for this variable, and add this new factor to the given model.