Design, Implementation, Application, and Dissemination of an Evolving Expression-Linkage Genetic Algorithms

摘要

AFOSR Grant No. F49620-97-I-0050 (Evolving Expression-Linkage GA) has come to a close and this report details project accomplishment. Particularly notable was the invention of a new competent GA, the Bayesian Optimization Algorithm (BOA) that works by budding and using a probabilistic model of the best solution points seen so far. BOA scales well and solves very hard problems quickly reliably and accurately. Another notable achievement was the completion of a Erick Cantu-Paz's magisterial PhD dissertation entitled 'Designing Efficient and Accurate Parallel Genetic Algorithms'. This thesis provides the first scaling laws that show why GAs are so easy to make parallel and how to do it better. These findings and accomplishments are discussed in the final project report together with others, including the development of a new theory of hybrid optimization and a. new theory of effective lime utilization. The report also details the affiliation and support of 26 researchers with the project, the publication, acceptance, or submission of 65 publications acknowledging AFOSR support, and numerous interactions and transitions.