Using SMILE > Discrete Bayesian networks > Discrete nodes and numeric domains

By default, discrete nodes represent discrete random variables that have categorical outcomes described by outcome identifiers. In SMILE 2, it is possible to model discrete random variables that are numerical in nature. The outcomes of the node can be associated with adjacent numeric intervals with explicitly specified borders, or with a set of discrete numeric values (one for each outcome). This new functionality is defined in the DSL_discDef class, which is a base class for all discrete node types.

There are four possible outcome types for discrete nodes:

•identifiers (the default)

•identifiers and numeric intervals

•numeric intervals with no identifiers

•identifiers and numeric point values

Nodes with intervals and point values are described in the next two subsections. See DSL_discDef reference for API details.