Introduction to Evolutionary Computing
Introduction to Evolutionary Computing
The gambler's ruin problem, genetic algorithms, and the sizing of populations
Evolutionary Computation
Dynamic populations in genetic algorithms
Proceedings of the 2008 ACM symposium on Applied computing
Dynamic hybrid fault models and the applications to wireless sensor networks (WSNs)
Proceedings of the 11th international symposium on Modeling, analysis and simulation of wireless and mobile systems
New approaches to reliability and survivability with survival analysis, dynamic hybrid fault models, and evolutionary game theory
AICI '09 Proceedings of the International Conference on Artificial Intelligence and Computational Intelligence
AICI '09 Proceedings of the International Conference on Artificial Intelligence and Computational Intelligence
International Journal of Bio-Inspired Computation
Chaotic populations in genetic algorithms
Applied Soft Computing
Hi-index | 0.03 |
In evolutionary computing (EC), population size is one of the critical parameters that a researcher has to deal with. Hence, it was no surprise that the pioneers of EC, such as De Jong (1975) and Holland (1975), had already studied the population sizing from the very beginning of EC. What is perhaps surprising is that more than three decades later, we still largely depend on the experience or ad-hoc trial-and-error approach to set the population size. For example, in a recent monograph, Eiben and Smith (2003) indicated: "In almost all EC applications, the population size is constant and does not change during the evolutionary search ." Despite enormous research on this issue in recent years, we still lack a well accepted theory for population sizing. In this paper, I propose to develop a population dynamics theory forEC with the inspiration from the population dynamics theory of biological populations in nature. Essentially, the EC population is considered as a dynamic system over time (generations ) and space (search space or fitness landscape ), similar to the spatial and temporal dynamics of biological populations in nature. With this conceptual mapping, I propose to 'transplant' the biological population dynamics theory to EC via three steps: (i ) experimentally test the feasibility--whether or not emulating natural population dynamics improves the EC performance; (ii ) comparatively study the underlying mechanisms--why there are improvements, primarily via statistical modeling analysis; (iii ) conduct theoretical analysis with theoretical models such as percolation theory and extended evolutionary game theory that are generally applicable to both EC and natural populations. This article is a summary of a series of studies we have performed to achieve the general goal [27][30]---[32]. In the following, I start with an extremely brief introduction on the theory and models of natural population dynamics (Sections 1 & 2). In Sections 4 to 6, I briefly discuss three categories of population dynamics models: deterministic modeling with Logistic chaos map as an example, stochastic modeling with spatial distribution patterns as an example, as well as survival analysis and extended evolutionary game theory (EEGT) modeling. Sample experiment results with Genetic algorithms (GA) are presented to demonstrate the applications of these models. The proposed EC population dynamics approach also makes survival selection largely unnecessary or much simplified since the individuals are naturally selected (controlled) by the mathematical models for EC population dynamics.