Understanding EA Dynamics via Population Fitness Distributions

It is clear from the study of complex non-linear systems in general, and evolutionary algorithms (EAs) in particular, that there is no single analysis tool or technique capable of providing a reasonably comprehensive understanding of the behavior of a complex non-linear system. Rather, we continue to develop sets of tools that collectively provide a more comprehensive view than any one tool. In this paper we provide an initial assessment of a new technique that is based on an old idea: the importance of understanding how the fitness distribution of an EA population changes over time.

• Type: Conference
• Year: 2003
• Where: GECCO
• Authors: Elena Popovici, Kenneth A. De Jong
• Keywords: Complex Non-linear System, EA Population Changes, GECCO 2003, Single Analysis Tool