Using SMILE Wrappers > Influence diagrams

Influence diagrams extend Bayesian networks by adding specialized node types to model decision-making:

•Decision nodes represent variables under the control of a decision maker, modeling the available alternatives explicitly as the possible outcomes of the node. They have no numerical parameters, only a discrete set of outcomes, and can be children of both decision and chance nodes. To create a decision node, pass DECISION to Network.add_node.

•Value nodes quantify the desirability of outcomes in the decision process. They are specified by the utility associated with each combination of parent node outcomes and can be children of decision and chance nodes. To create a value node, use UTILITY with Network.add_node.

•Multi-attribute utility (MAU) nodes combine multiple value nodes to define a multi-attribute utility function. The function can be specified as a set of weights for a linear function—making the node an additive linear utility (ALU)—or as expressions that reference parent value node identifiers. MAU nodes can be children of decision, value, and other MAU nodes. If decision parents exist, the MAU definition includes a separate set of weights or expressions for each combination of decision parent outcomes. To create a MAU node, pass MAU to Network.add_node. By default, MAU nodes are defined by weights; to use expressions, call Network.set_mau_expressions. For a list of available functions and syntax, see the Equations reference chapter.

As in Bayesian networks, the results of influence diagram inference are stored in node values and accessed via Network.get_node_value. However, interpretation is extended for influence diagrams. The returned matrices are indexed by indexing parents, which are unobserved decision nodes that precede the current node or unobserved chance nodes that are predecessors of decision nodes and should have been observed before decisions can be made. Use Network.get_indexing_parents to retrieve their handles or Network.get_indexing_parent_ids to retrieve their identifiers. The set of outcomes of indexing parents is called a policy.

After successful inference in an influence diagram, node values represent:

•Chance and deterministic nodes: posterior probabilities for each policy

•Decision nodes: expected utilities for all outcomes and for each policy

•Value and MAU nodes: expected utility for each policy

See Tutorial 4 for a simple influence diagram demo program.