•  
  •  
 

Scholarly Horizons: University of Minnesota, Morris Undergraduate Journal

Document Type

Article

Abstract

In this paper, we will provide an overview of the paper “Phylogeny-informed fitness estimation for test-based parent selection” by Lalejini, et al. [7] Phylogenies, or ancestry trees, provide a detailed look into the evolutionary journey of a population. In evolutionary computation, a phylogeny can represent the progress of an evolutionary algorithm through a search space. Although phylogenetic analysis is mainly used to deepen the understanding of evolutionary algorithms after they have been run, this study explores its potential use in real-time to enhance parent selection during evolutionary searches. The research by Lalejini, et al. introduces the concept of phylogeny-informed fitness estimation, leveraging a population’s phylogeny to predict fitness values. This method is tested using both down-sampled lexicase and cohort lexicase selection algorithms across four genetic programming (GP) problems. The findings suggest that using phylogenies to estimate fitness values can improve the performance of down-sampled lexicase selection, fostering better diversity and search space exploration.

Primo Type

Article

Download

Share

COinS