Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (14): 80-84.

• 理论研究 • Previous Articles     Next Articles

Evaluating genes without expression trees construction

LI Chuan1,TANG Chang-jie1,CHEN Yu1,DAI Shu-cheng1,QIU Jiang-tao1,LUO Qian1,ZHU Jun2   

  1. 1.School of Computer,Sichuan University,Chengdu 610065,China
    2.Birth Defects Supervising Centre,Chengdu 610065,China
  • Received:2008-01-18 Revised:2008-03-25 Online:2008-05-11 Published:2008-05-11
  • Contact: LI Chuan


李 川1,唐常杰1,陈 瑜1,代术成1,邱江涛1,罗 谦1,朱 军2   

  1. 1.四川大学 计算机学院,成都 610065
    2.中国出生缺陷监测中心,成都 610065
  • 通讯作者: 李 川

Abstract: Traditional ET(Expression Tree)-based GEP has a major performance defect:Repetitive ET traverses and calculations.This paper proposes a novel model called Scale-Based GEP to solve the problemVariable matrix is used to avoid repetitive calculations and traversing in ET-based gene evaluation.Experiments show that Scalebased evaluation outperforms ET-based method 3~5 times constantly.

Key words: data mining, gene expression programming, Scale-based gene expression

摘要: 传统基于表达式树ET的基因评估从性能角度讲主要缺点是:重复遍历表达式树和进行大量重复计算。提出基于Scale的基因评估。变量矩阵用来避免基因评估中的重复计算。实验表明,在绝大多数数据分布下和参数选择情况下:基于Scale的基因评估较基于 ET 的基因评估快 3~5 倍。

关键词: 数据挖掘, 基因表达式编程, 基于Scale的基因表达