Title:  Creating Human-competitive Algorithms using Multiple Objective Vector-based Genetic Programming

Committee: 

Dr. T. Michaels, Advisor        

Dr. Bennett, Chair

Dr. J. Michaels

Abstract: The objective of the proposed research is to increase the Georgia Tech Multiple Objective Evolutionary Programming (GTMOEP) framework's capabilities by seeking methods to reduce compute time and produce solutions in fewer evaluations. In order to reduce the number of individuals needed to reach optimal solutions we seek methods that will increase the viability of offspring. The ideas proposed here include the concept of a fuzzy Pareto front, identifying species in the population, strategic selection of data for evolution, and techniques for working with larger numbers of objectives. We would also like to demonstrate performance on a time domain or spatial domain problem, showcasing GTMOEPs capability to create and leverage features from higher dimensional data.