Choice, information and computing
Abstract
This thesis investigates the relationship among choice, information and computing. In particular, an economic decision model is used to construct optimal algorithms and conditions for when and optimization approach is appropriate are presented. The existence of a linear ordering over the experiment set enables us to use dynamic programming to construct an optimal algorithm for the file search problem. An explanation of the linear ordering condition and why it enables us to use dynamic programming, as well as the dynamic programming formulation are presented. The common index problem is a generalization of the file search problem. In the common index problem, the user wants to find the best available record rather than a particular record. Under certain conditions we can formulate the common index problem so that there is a linear ordering over the set of experiments. We present these conditions, the dynamic programming formulation and sufficient conditions for obtaining complete information. An optimal strategy for a variant of the common index problem where utility is a function of multiple attributes is also presented. The strategy resembles that used in some psychological choice models. The relationship between these models and the decision model used in this thesis are also discussed. Finally, the use of composite experiments is examined and their relationship to parallel and distributed processing is explained. We present sufficient conditions for when dynamic programming can be used to construct optimal algorithms when the experiments are constructed from primary experiments.
Degree
Ph.D.
Advisors
Whinston, Purdue University.
Subject Area
Management
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.