基于MHC调控的免疫公式发现算法

MHC Regulation Based Immune Formula Discovering Algorithm

在分析了基于遗传原理的公式发现方法的优势与不足的基础上,根据免疫原理和MHC(major histocompatibility complex)在免疫系统中的调控作用,提出了一种应用于公式发现领域的算法IFDA(immune formula discovering algorithm)来解决公式进化中优良结构不易保护的问题.该算法将公式翻译成树状图,并按深度优先的编码方法形成抗体的恒定区和可变区代码,把公式片段编码成为MHC代码,借鉴MHC调控原理指导抗体进化,寻找出数据集合中蕴涵的规律,并用公式的形式表示.通过对多组基准数据的实验说明,此方法在公式复杂度和收敛速度方面比基因表达式算法有更好的性能.

Aiming at the difficulty in good segments of the formula to be inherited in formula discovering using gene expression programming （GEP）, this paper proposes an innovative immune formula discovering algorithm （IFDA）, which is actually inspired by MHC （major histocompatibility complex） regulation principle of immune theory. In IFDA, the formula are mapped as a tree structure and transformed into both constant and variation section of antibody with a depth-first mechanism while its fragment is encoded into the MHC. By the feature of MHC regulation, IFDA mines the dataset to discover the proper formula very quickly. Many data are benchmarked for verifying the performance of IFDA in which all results from experiments show that the IFDA can really provide better performance than GEP in both convergence speed and formula complexity.