一种基因表达式程序设计的解码方法  

A Decodingmethod for Gene Expression Programming

作　　者：郭勇[1] 何锫[2,3] 胡洋[1] 李明飞[1] 

机构地区：[1]黔南民族师范学院计算机科学系,贵州都匀558000 [2]广州大学计算机科学与教育软件学院,广州510006 [3]武汉大学软件工程国家重点实验室,武汉430072

出　　处：《小型微型计算机系统》2016年第2期354-360,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61170199)资助;贵州省科技厅联合基金项目(20147440;20157727)资助;贵州省教育厅教学质量工程重点项目(2012426)资助

摘　　要：基因表达式程序设计(GEP)的解码通常仰赖表达式树的建立和后序遍历技术,因而解码复杂度、性能自然成为GEP应用的要害所在.在分析GEP基因型与表现型关系的基础上,提出一种称谓RL-GEP的新型解码方法.新方法基于0目操作符概念、工程应用与系统设计的原则,采用"一次读码多样本解析"和直接对线性编码的基因型实施解码等方法来提高解码效率,算法模型简单修改即可得到一种新型的传统GEP"无树解码"方法,具有良好的扩展性.RL-GEP不仅与传统GEP具有相同的表达能力与表现型空间,而且易于理解、应用和扩展.从求解回归问题的实验看来,本方法和经典GEP有相似问题求解的能力,但效率更高.Decoding complexity and performance are of important concerns to applications of GEP （ Gene Expression Programming ）, which generally decodes chromosomes based both on the construction of expression tree and post-order traversal. Taking into consider- ation the relationship between genotype and phenotype, a novel GEP decoding method, called RL-GEP, is proposed for interpreting genes in a straightforward manner in light of concept of 0-ary operator as well as principles of engineering applications and system de- signs. Genotype decoding methods such as ＇An Analysis of Reading-Code Mutl-Sample＇ and ＇Direct Linear Coding＇ are adopted to im- prove the efficiency of decoding. A novel traditional GEP ＇Non-Expression Tree＇ method can be produced through a simple modifica- tion of arithmetic model, which is of better expansibility and universality. RL-GEP not only possesses the same expressiveness and phenotypic space as that of GEP, but also is easy to understand and use, and convenient to be extended. Experimental results demon- strate that RL-GEP has similar capability to solve regression problems as traditional GEP, but is more efficient than the latter.

关 键 词：基因表达式程序设计 基因码 解码 符号回归 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]