Master of Engineering Science
Electrical and Computer Engineering
Dr. Abdallah Shami and Dr. Serguei Primak
This thesis examines two topics from the field of computational optimization; architectural layout generation and parallel linear programming. The first topic, a modern problem in heuristic optimization, focuses on deriving a general form of the optimization problem and solving it with the proposed Evolutionary Treemap algorithm. Tests of the algorithm's implementation within a highly scalable web application developed with Scala and the web service framework Play reveal the algorithm is effective at generated layouts in multiple styles. The second topic, a classical problem in operations research, focuses on methodologies for implementing the Simplex Algorithm on a parallel computer for solving large-scale linear programming problems. Implementations of the algorithm's data-parallel and task parallel forms illuminate the ideal method for accelerating a solver. The proposed Multi-Path Simplex Algorithm shows an average speed up of over two times that of a popular open-source solver, showing it is an effective methodology for solving linear programming problems.
de Vlugt, Bradley J., "Modern Optimization Algorithms and Applications: Architectural Layout Generation and Parallel Linear Programming" (2015). Electronic Thesis and Dissertation Repository. 3208.
Available for download on Friday, September 01, 2017