Abstract

Evolutionary algorithms have shown substantial progress in recent years, especially in neural architecture search and neuroevolution applications. Despite their effectiveness, analyzing and understanding the evolutionary paths these algorithms traverse to reach solutions remains challenging. These algorithms often involve distributed computing strategies, which can include subpopulations or islands, and they explore massive or even unbounded search spaces in both continuous and non-continuous domains. Manually examining individual solutions to understand the evolutionary dynamics is often infeasible due to large population sizes, large genome sizes, and high generation counts. This work introduces a new methodology for visualizing neuroevolution population dynamics called genetic distance projections, along with a novel neural network-based method for generating these representations. This methodology is evaluated empirically and is found to perform better than other, more traditional methods in generating these representations. The usefulness of this methodology was validated with three different neuroevolution frameworks, including NEAT, EXAMM, and an experimental one, EXA-STAR. In addition, the debugging potential of this methodology is demonstrated on EXAMM.

Library of Congress Subject Headings

Evolutionary computation; Genetic algorithms; Neural networks (Computer science)

Publication Date

5-2-2025

Document Type

Thesis

Student Type

Graduate

Degree Name

Artificial Intelligence (MS)

Department, Program, or Center

Information, School of

College

Golisano College of Computing and Information Sciences

Advisor

Travis Desell

Advisor/Committee Member

Jamison Heard

Advisor/Committee Member

Tamas Wiandt

Additional Files
genetic distance projection visuals.zip (119081 kB)
Supplement
patterson_evan_announcement.pdf (92 kB)
Supplement 1
patterson_evan_proposal.pdf (88 kB)
Supplement 2

Campus

RIT – Main Campus

Plan Codes

AI-MS

Download

Additional files available below

Share

COinS