Session:

LBP - Late Breaking Papers

Title:

An Adaptive Diploid Evolutionary Algorithm for Floating-Point Representations in Dynamic Environments

Authors:

A. Sima Uyar

Abstract:

Many flavors of different EA approaches have been suggested in literature to address various requirements that arise when the environment is dynamic. One such line of research focuses on using memory-based techniques. However, their use is very limited and they are only considered to be advantageous when the environment periodically returns to previously visited states. In addition, most of these approaches require a change detection mechanism to adapt and their performance partly depends on an accurate detection mechanism. In this study, an implicit memory-based approach for floating-point representations is introduced.. A technique similar to estimation of distribution algorithms is used to automatically adapt the population to the change. This approach aims at addressing some of the deficiencies reported for memory-based approaches and it requires no change detection mechanism to trigger the adaptation process. The Moving Peaks Benchmark is used in the experiments to show the desired properties of the proposed approach. Preliminary experimental results are promising and promote further study.