Differentiate coevolutionary algorithm

作者：Li, Bi^*; Zhao, Xiao-Fang; Zhang, Qian-Sheng; Tan, Shi-Heng

来源：Journal of Convergence Information Technology, 2011, 6(4): 180-187.

DOI：10.4156/jcit.vol6.issue4.21

• Benchmark functions
• Co-evolution
• Co-evolutionary algorithm
• Convergence
• Dynamic environments
• Exploitation and explorations
• Faster convergence
• Fitness

This paper describes a Differentiate CoEvolutionary Algorithm (DCEA), in which each individual has not only chromosomal fitness but also survival fitness. The chromosomal fitness is determined by the individual's chromosome completely, which is the same as that in the genetic algorithm (GA). The survival fitness is determined by the individual's chromosome and the coevolutionary relationships in which it is involved, measuring the individual's survival ability. DCEA embodies all the coevolutionary relationships ubiquitous in natural ecosystems. The total influence received by an individual from all the other individuals has only three possibilities: benefit, harm or neutral. DCEA is tested on two numeric benchmark function optimization problems. Experimental results show that DCEA outperforms GA, showing faster convergence performance and obtaining a better balance between exploitation and exploration. But the value of coevolutionary factors should adjust self-adaptively according to the optimized objects or the dynamic environments.

• 出版日期2011-4
• 单位华南理工大学; Cisco