Grammar Based Genetic Programming Using Linear Representations

• 作者: LUYinan WANGFei ZHANGHong
• 来源期刊: 电子学报(英文版)
• 学科: 工学
• 关键词: 遗传规划 解析树 线性表示 CFGs 计算机
• 年，卷（期）: 2003,（1）
• 页码范围: 75-78
• 页数: 4页
• 分类号: TP301.6
• 语种: 中文

摘要

In recent years,there has been a great interest in genetic programming(GP),which is used to solve many applications such as data mining,electronic engineering and pattern recognition etc.. Genetic programming paradigm as a from of adaptive learning is a functional approach to many problems that require a nonfixed representation and GP typically operates on a population of parse which usually represent computer programs whose nodes have single data type.In this paper GP using context-free grammars(CFGs) is described.This technique separates search space from solution space through a genotype to phenotype mapping.The genotypes and phenotypes of the individuals both act on different linear representations.A phenotype is postfix expression,a new method of representing which is described by making use of the definition and related features of a context-free grammar,i.e.a genotype is a variable length,linear valid genome determined by a simplifled derivation tree(SDT) generated from a context-free grammar.A CFG is used to specify how the possible solutions are created according to experiential knowledge and to direct legal crossover(ormutation)operations without any explicit reference to the process of program generation and parsing,and automatically ensuring typing and syntax correctness.Some related definitions involving genetic operators are described.Fitness evaluation is given.This technique is applied to a symbol regression problem-the identification of nonlinear dynamic characteristics of cushioning packaging.Experimental results show this method can flnd good relations between variables and is better than basic GP without a grammar.Future research on it is outlined.