Contemporary Artificial Intelligence

Chapter 8

Decision Analysis

In general, the information obtained by doing inference in a Bayesian network

can be used to arrive at a decision, even though the Bayesian net work itself

does not recommend a decision. In this chapter, we extend the structure of a

Bayesian network s o that the network actually does recommend a decision. Such

a network is called an influence diagram. Section 8.1 introduces decision trees,

which are mathematically equivalent to influence diagrams, but whic h have dif-

ficulty representing large instances because their size grows exponentially with

the number of variables. In Section 8.2 we discuss influence diagrams, whose

size ...