Accession Number : ADA290856
Title : Selecting Optimal Experiments For Feedforward Multilayer Perceptrons.
Descriptive Note : Doctoral thesis,
Corporate Author : AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH
Personal Author(s) : Belue, Lisa M.
PDF Url : ADA290856
Report Date : MAR 1995
Pagination or Media Count : 187
Abstract : Where should a researcher conduct experiments to provide training data for a multilayer perceptron? This question is investigated and a statistically-based method for optimally selecting experimental design points for multilayer perceptrons is introduced. Specifically, a criterion is developed based on the size of an estimated confidence ellipsoid for the weights in the multilayer perceptron. This criterion is minimized over a set of exemplars to find optimal design points. Initially, single output networks are examined. An example is used to demonstrate the superiority of optimally selected design points over randomly chosen points and points chosen in a grid pattern. Also, two measures are successfully used to rank the design points in terms of their importance. Two methods are presented to significantly reduce complexity-a distributed linear feedthrough network structure and a weight subset method. Next, multiple output networks are examined. The criterion in this framework becomes more complex and a simplifying technique is employed to judiciously choose desired outputs of the network to produce uncorrelated actual outputs. Finally, the methods described above are integrated and tested on two applications dealing with aircraft survivability. In both cases, simulating the indicated experiments produced a superior multilayer perceptron. (AN)
Descriptors : *NEURAL NETS, *PATTERN RECOGNITION, ALGORITHMS, OPTIMIZATION, DATA MANAGEMENT, DISTRIBUTED DATA PROCESSING, EXPERIMENTAL DESIGN, RANDOM VARIABLES, GRIDS, LEARNING MACHINES, THESES, INPUT OUTPUT PROCESSING, ESTIMATES, REGRESSION ANALYSIS, KNOWLEDGE BASED SYSTEMS, NONLINEAR ANALYSIS, COVARIANCE, DESIGN CRITERIA, CONFIDENCE LEVEL, DISCRIMINATE ANALYSIS.
Subject Categories : Cybernetics
Distribution Statement : APPROVED FOR PUBLIC RELEASE