COMING SOON! PQDT Open is getting a new home!

ProQuest Open Access Dissertations & Theses will remain freely available as part of a new and enhanced search experience at www.proquest.com.

Questions? Please refer to this FAQ.

Dissertation/Thesis Abstract

A Maximin Optimization Method for Improved System Performance
by Liaghati, Hassan Robert, Ph.D., The George Washington University, 2021, 124; 28410744
Abstract (Summary)

Complex systems are challenging for systems engineers to design, due to their potential for strong emergent behavior. A significant source of emergent behaviors for modern systems are complex operating environments. This research proposes a Maximin Optimization Method to design a system for the set of operating environments or conditions in which the system is required to operate. The research developed two implementations of the Maximin Optimization Method: the Co-Evolutionary Maximin Optimization Method and the Nested Maximin Optimization Method. This research provides a description of the two methods and steps for their implementation. Two example systems are then proposed and optimized using the respective Maximin Optimization Methods. The first example system, an Air Defense System, is optimized using the Co- Evolutionary Maximin Optimization Method and the resulting system configuration is compared to two benchmark system configurations. The second example system, a ResNet-18 based image classifier is optimized using the Nested Maximin Optimization Method and the resulting system configuration is compared to eight benchmark system configurations. In both examples, the system design resulting from the Maximin Optimization Method demonstrated significantly improved performance over the benchmarks as demonstrated by 10,000 Monte Carlo runs. Both example systems resulted in more consistent performance across the range of required operating environments, while either improving or maintaining average performance, when compared to the benchmark configurations.

Indexing (document details)
Advisor: Mazzuchi, Thomas A., Sarkani, Shahram
Commitee: Etemadi, Amir, Holzer, Thomas H., Blackford, Joseph P.
School: The George Washington University
Department: Systems Engineering
School Location: United States -- District of Columbia
Source: DAI-B 82/10(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Systems science, Applied Mathematics, Artificial intelligence
Keywords: Image classification, Mathematical programming, Maximin optimization, Resilience, Systems engineers
Publication Number: 28410744
ISBN: 9798597096179
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest