关键词: 基因表达式编程, 局部最优, 个体多样性, 自适应进化辈数

Abstract: Gene expression programming （GEP） is a powerful tool for function discovery. However， it may suffer from prematurity in solving problems. The existing methods for preventing GEP from prematurity usually embark after the occurrence of prematurity and hence are inefficient. This paper proposes a novel strategy for pre-monitoring the premature of GEP. The main contributions include： （1） By evaluating the diversity of chromosomes， an individual selection mechanism is proposed. And the diversity contribution-GEP （DC-GEP） compromising fitness maximization and diversity maintenance is designed； （2） A self-determined stop criterion， which is incorporated into the adaptive DC-GEP， is proposed； （3） The experimental results indicate that the proposed algorithm DC-GEP can enhance the evolution efficiency of mining complex functions by 70% compared with the basic GEP.

Key words: gene expression programming （GEP）, local optimization, individual diversity, adaptive generation