Inference

To exchange information over the tree factor graph, the FactorGraph provides forward–backward GBP algorithm. We advise the reader to read the Section forward–backward message passing schedule which provides a detailed description of the inference algorithm.

Each of the inference functions accepts only the composite type ContinuousTreeModel, i.e., an output variable of the function gbp = continuousTreeModel().


Forward inference

The set of functions that can be used to preform forward message inference:

forwardVariableFactor(gbp)
forwardFactorVariable(gbp)

Backward inference

The set of functions that can be used to preform backward message inference:

backwardVariableFactor(gbp)
backwardFactorVariable(gbp)

Marginal inference

To compute marginals the FactorGraph provides the function:

marginal(gbp)

Same as before, the function accepts the composite type ContinuousTreeModel.