The Genetic Algorithm (GA) is a popular global search algorithm that enjoyed great success applications in many fields. It’s very flexible and effective. However, GA is hindered by performance challenges: difficult-to-reach optimal solutions for complex problems and slow convergence speed for difficult problems.

In this dissertation, we introduce the shadow price concept into GA and propose a new two-measurement GA. The new algorithm uses fitness values to measure overall solutions and shadow price to evaluate components.  Shadow price is used to compare components, evaluate their potential contributions, and set evolution direction. With shadow price guiding the operators, the GA can generate better solutions and speed up the optimization process.

The proposed shadow price concept complements the fitness evaluation in the GA’s search process. Fitness values are used to compare and filter solutions. Shadow prices are used to compare and select components in the search process. Together, they constitute the proposed two-measurement GA. The proposed new search algorithm was applied to many difficult problems and received very good results. Experimental results validated the new algorithm’s improvement on solution quality and convergence speed.

Committee
Dr. Yanqing Zhang (chair)
Dr. Yingshu Li
Dr. Raj Sunderraman
Dr. Yichuan Zhao