Date of Award
Fall 2007
Document Type
Dissertation - Restricted
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering
First Advisor
Setiz, Martin A.
Second Advisor
Corliss, George F.
Third Advisor
Heinen, James
Abstract
Historically academic exercises train neural networks with hundreds of cases. In a manufacturing environment that number is in the thousands. In this dissertation, however, our database has numbers in the tens - magnitudes lower than conventional training sets. There are a number of situations that have a very small database, but the possibility of expanding it is either out of your control, too costly, or not feasible. In such situations, you have to work with the data available and extract meaningful information. This dissertation presents a method that extracts meaningful information from a phenotypic database of a genetic animal study from the Medical College ofWisconsin. The goal is to develop a method for extracting information from the very small database that is representative of the distinguishing parameters (phenotypes) relative to diseased (hypertensive) or non-diseased (normal) genetic rats for a particular genetic configuration. The phenotypic information obtained through laboratory experiments is done on a specific genetic configuration of the animal (rat). This phenotypic information provides the foundation for determining the biological aspects of the specific disease being studied, in this case the multi-genetic disease of hypertension...