Date of Award
Master of Science
Ajay Gupta, Ph.D.
Elise de Doncker, Ph.D.
Jeffrey Horn, Ph.D.
Cooperative evolution, fitness sharing, niche genetic algorithm, parallel optimization, resouce-defined fitness sharing, resource sharing
Masters Thesis-Open Access
The exploitation of niches by genetic algorithms (GAs) is a computationally expensive, but effective, methodology for solving complex open problems and real-world applications. Niching, differentiated on the modality of sharing, casts problems in terms of the specific resources available. These concepts arise from the broader natural algorithms that encapsulate the ideas and theories used in artificial intelligence. In remediating the computational costs, a study on exploiting niche-defined parallel structures is performed in the contest of the resourcedefined fitness sharing (RFS) algorithm.
Sharing is a natural algorithm paradigm that emulates the use of resources within an environment or population. Defining these resources presents two closely related techniques: resource and fitness sharing. Resource sharing seeks to apply the finite resources available, whereas fitness sharing assesses a population based on merit.
Resource and fitness sharing exhibit a duality within the sharing paradigm. Consequently, problems that have resource-defined niches are incompatible with fitness sharing. Conversely, resource sharing has great difficulty in managing the non-linear interactions among shared fitnesses. The RFS algorithm was developed to resolve these deficiencies and is the focus of this study for the parallel optimization of niching structures.
Rogers, Blayne A., "Parallel Resource Defined Fitness Sharing: A Study on Parallel Optimizations for Niching Algorithms" (2022). Masters Theses. 5327.