CAJ | 학술논문

传统的演化规划（CEP）依赖高斯变异算子，而快速演化规划（FEP）选择柯西分布作为主要的变异算子。改进的快速演化规划（IFEP）是将柯西变异算子和高斯变异算子的搜索倾向混合起来。每个父代生成两个子代，一个有柯西变异算子，另一个有高斯变异算子，然后比较这两个子代，将变现好的一个保留作为下一代。在本文，我们提出了一种极值导向的自适应变异算子演化规划（OSDEP）,它的基本思想是将当前最优搜索方向引入柯西变异算子中，在OSDEP中每个个体在柯西变异算子作用下,再沿着当前最优解的方向进行搜索。大量的数值试验对OSDEP, IFEP, FEP和CEP进行比较。从这些具有广泛代表性的七个测试函数的数值试验结果，我们可以观察到对于单峰函数、有少数局部最优的多峰函数和有很多局部最优的多峰函数DSEP比IFEP, FEP和CEP都要表现好。
전통적연화규화（CEP）의뢰고사변이산자，이쾌속연화규화（FEP）선택가서분포작위주요적변이산자。개진적쾌속연화규화（IFEP）시장가서변이산자화고사변이산자적수색경향혼합기래。매개부대생성량개자대，일개유가서변이산자，령일개유고사변이산자，연후비교저량개자대，장변현호적일개보류작위하일대。재본문，아문제출료일충겁치도향적자괄응변이산자연화규화（OSDEP）,타적기본사상시장당전최우수색방향인입가서변이산자중，재OSDEP중매개개체재가서변이산자작용하,재연착당전최우해적방향진행수색。대량적수치시험대OSDEP, IFEP, FEP화CEP진행비교。종저사구유엄범대표성적칠개측시함수적수치시험결과，아문가이관찰도대우단봉함수、유소수국부최우적다봉함수화유흔다국부최우적다봉함수DSEP비IFEP, FEP화CEP도요표현호。
The Classical Evolutionary Programming (CEP) relies on Gaussian mutation, whereas Fast Evolutionary Programming (FEP) selects Cauchy distribution as the primary mutation operator, Improved Fast Evolutionary (IFEP) selects the better Gaussian and Cauchy distribution as the primary mutation operator. In this paper, we propose a self-adaptive Evolutionary Programming base on Optimum Search Direction (OSDEP) in which we introduce the current best global individual into mutation to guide individuals to converge according to the global search direction. Extensive empirical studies have been carried out to evaluate the performance of OSDEP, IFEP, FEP and CEP. From the experimental results on seven widely used test functions, we can show that OSDEP outperforms all of IFEP, FEP and CEP for all the test functions.