基于GEP多项式函数关系分解.pdfVIP

基于GEP 的多项式函数关系分解1 汪锐 唐常杰 段磊 陈宇 四川大学 计算机学院 {wangrei, tangchangji e }@ 摘要：为使从实验科学数据中发现的函数关系更加直观实用，此文利用GEP 方法实现了多项式函 数分解。本文主要工作如下：（1）提出了基于基因表达式编程GEP 的多项式函数分解方法 GPF(GEP Polynomial Factorization) 。能将任意多项式函数关系，包括实数域上不可分的多项 式，按指定精度分解若干个低次多项式函数之积。(2) 突破了传统的多项式分解方法，采用 数据挖掘的思想，利用基因表达式编程GEP 技术提出了GPF 算法，（3 ）GPF 中采用了有 特色的概率相关因子优化对GEP 中的适应度函数进行优化，使得精度提高了27% 。（4 ）GPF 提出了宽松环境进化策略LEE （Loose Environment Evolution ）使得GEP 成功率比传统技术 提高了最大58 倍。（5 ）将GPF 推广到观察数据集上的多个函数关系挖掘MFM （Multiple Function Mining ）。（6 ）通过一系列的实验证明了GPF 的有效性。 关键字：多项式函数分解，概率相关因子，宽松环境进化，基因表达式编程 Polynomial functions factorization based on GEP WANG Rui, TANG Changji e, DUAN Lei, CHEN Yu Abstract: In order to discover more interesting and practical functions from experimental scientific data, this article proposes an approach to factorize polynomial functions based on GEP(Gene Expression Programming). The main contributions are as follows: ( 1) Proposing an approach named GPF(GEP Polynomial Factorization) for polynomial functions factorization based on GEP. It can discover all factors with specified precision for arbitrary polynomial function, including impartible polynomial function over real number field. (2) Presenting the GPF algorithm based on GEP techniques in spite of the limits of traditional factorization methods. (3) Optimizing the fitness function in GEP by special approach called probability correlation factor, which increases the precision by 27%. (4) Adopting a brand new strategy named LEE （Loose Environment Evolution ）to improve the success-probability by 58 times compared with traditional approaches.(5) Promoting GPF approach to MFM （Multiple Function Mining ）on observed data set. (6) By extensive experiments, demonstrating the power