ADM 2302 Lecture Notes - Lecture 17: Expected Value Of Perfect Information, Decision Tree Learning

• School: University of Ottawa
• Department: Administration
• Course: ADM 2302
• Professor: Jonathan Yu- Meng L I

Document Summary

> still evpi acts as an upper bound on the value of imperfect information. > simple way of examining trade-o s between various potential decisions. > evpi lets you know how valuable additional information is to you. > works well if a) there is a limited number of decisions options b) there is a limited number of states of nature and c) there is only one-o decision to make. When there is more than one decision to be made . > so far, the decisions we have studied have been quite simple, usually consisting of a single decision and a single uncertain event. > we"ll now show how to solve more complex decision problems, by repeatedly using the principles we have developed for simple problems. > the basic tool we will use to structure complex problems is the decision tree. > the major new principle used in solving complex decisions is the idea of backward induction.