Knowledge-based approaches to self-adaptation in cultural algorithms
Chan‐Jin Chung · 1997
Cultural Algorithms are computational self-adaptive models which consist of a population and a belief space. Problem solving experience of individuals selected from the population space by the acceptance function is generalized and stored in the belief space. This knowledge can then control the evolution of the population component by means of the influence function. Here, we examine the role that different forms of knowledge can play in the self-adaptation process for evolution-based function optimizers. In particular, we compare various approaches using normative and situational knowledge in guiding the search process. Also we investigate the impact of different acceptance and influence functions on the system's performance by employing both static and flexible fuzzy approaches. Evolutionary Programming is used to implement the population space. The best performance is produced using knowledge to decide both step size and direction in most cases. In addition, the use of a fuzzy acceptance and influence function appears to be a promising one. All the results in this study exhibit that Cultural Algorithms are a naturally useful framework for self-adaptation and that the use of a cultural framework to support self-adaptation in Evolutionary Programming can produce substantial performance improvements as expressed in terms of (1) system success ratio, (2) execution CPU time, and (3) convergence (mean best solution) for a given set of function minimization problems. The nature of these improvements and the type of knowledge that is most effective in producing them depends on the structure of the problem. While in most cases, the best performance is produced using knowledge to decide both step size and direction, there are situations where controlling only the direction or the step size produces the best results. Also normative knowledge appears to be the dominant and general purpose knowledge source for the optimization functions here.