Abstract:
A key challenge in genetic and evolutionary computation (GEC) research is the design of competent genetic algorithms (GAs). By competent we mean GAs that can solve hard problems, quickly, reliably, and accurately, and much progress has been made along these lines (see for example, Dave Goldberg’s Design of Innovation). In essence, competent GA design takes problems that were intractable with first generation GAs and renders them tractable, oftentimes requiring only a subquadratic number of fitness evaluations. However, for large-scale problems, the task of computing even a subquadratic number of function evaluations can be daunting. This is especially the case if the fitness evaluation is a complex simulation, model, or computation. This places a premium on a variety of efficiency enhancement techniques. While competence leads us from intractability to tractability, efficiency enhancement takes us from tractability to practicality.
In this tutorial, we will consider a four part decomposition of efficiency-enhancement techniques: (1) Parallelization, (2) Hyrbridization, (3) Time Continuation, and (4) Evaluation Relaxation. We will develop a principled design methodology for different methodologies in each of efficiency-enhancement technique categories using facetwise modeling and dimensional arguments. The principled design methodology not only enables us to predict the maximum speed-up and scalability of each of the efficiency-enhancement methods, but also yields practical guidelines of using them in real-world problems.