Doctoral Dissertations
Abstract
"A linear acceleration technique, LAT, is developed which is applied to three conjugate direction algorithms: (1) Fletcher-Reeves algorithm, (2) Davidon-Fletcher-Powell algorithm and (3) Grey's Orthonormal Optimization Procedure (GOOP). Eight problems are solved by the three algorithms mentioned above and the Levenberg-Marquardt algorithm. The addition of the LAT algorithm improves the rate of convergence for the GOOP algorithm in all problems attempted and for some problems using the Fletcher-Reeves algorithm and the Davidon-Fletcher-Powell algorithm. Using the number of operations to perform function and derivative evaluations, the algorithms mentioned above are compared. Although the GOOP algorithm is relatively unknown outside of the optics literature, it was found to be competitive with the other successful algorithms. A proof of convergence of the accelerated GOOP algorithm for nonquadratic problems is also developed"--Abstract, page ii.
Advisor(s)
Rigler, A. K.
Committee Member(s)
Reisbig, R. L.
Plummer, O. R.
Gillett, Billy E.
Engelhardt, Max
Department(s)
Mathematics and Statistics
Degree Name
Ph. D. in Mathematics
Publisher
University of Missouri--Rolla
Publication Date
1972
Pagination
vii, 95 pages
Note about bibliography
Includes bibliographical references (pages 71-75).
Rights
© 1972 Larry Wilmer Cornwell, All rights reserved.
Document Type
Dissertation - Open Access
File Type
text
Language
English
Subject Headings
Mathematical optimizationRegression analysisNonlinear theories
Thesis Number
T 2630
Print OCLC #
6038898
Electronic OCLC #
878077966
Recommended Citation
Cornwell, Larry Wilmer, "An acceleration technique for a conjugate direction algorithm for nonlinear regression" (1972). Doctoral Dissertations. 1861.
https://scholarsmine.mst.edu/doctoral_dissertations/1861