基于概率分布的多峰演化算法

发布时间：2019-06-17 11:21

【摘要】：演化算法通过模拟自然界生物迭代演化的智能现象来求解优化问题,因其不依赖于待解问题具体数学模型特性的优势,已成为求解复杂优化问题的重要方法.分布估计算法是一类新兴的演化算法,它通过估计种群中优势个体的分布状况建立概率模型并采样得到子代,具有良好的搜索多样性,且能通用于连续和离散空间的优化问题.为进一步推动基于概率分布思想的演化算法发展,概述了多峰优化演化算法的研究现状,并总结出2个基于概率分布的演化算法框架:面向多解优化的概率分布演化算法框架和基于概率分布的集合型离散演化算法框架.前者针对现有的演化算法在求解多峰多解的优化难题时缺乏足够的搜索多样性的缺点,将广义上基于概率分布的演化策略与小生境技术相结合,突破多解优化的搜索多样性瓶颈;后者围绕粒子群优化等部分演化算法在传统上局限于连续实数向量空间的不足,引入概率分布估计的思想,在离散的集合空间重定义了算法的演化操作,从而提高了算法的可用性.
[Abstract]:The evolutionary algorithm solves the optimization problem by simulating the intelligent phenomenon of the iterative evolution of natural organisms. Because it does not depend on the advantages of the specific mathematical model of the problem to be solved, it has become an important method to solve the complex optimization problem. Distribution estimation algorithm is a new kind of evolutionary algorithm. It establishes probability model by estimating the distribution of dominant individuals in the population and samples to obtain offspring. It has good search diversity and can be applied to continuous and discrete space optimization problems. In order to further promote the development of evolutionary algorithm based on probability distribution, the research status of multi-peak optimization evolution algorithm is summarized, and two evolutionary algorithm frameworks based on probability distribution are summarized: probability distribution evolution algorithm framework for multi-solution optimization and set discrete evolution algorithm framework based on probability distribution. In view of the shortcomings of the existing evolutionary algorithms in solving multi-peak and multi-solution optimization problems, the former combines the evolution strategy based on probability distribution with niche technology in a broad sense to break through the bottleneck of search diversity in multi-solution optimization. The latter is traditionally limited to the shortcomings of continuous real vector space around particle swarm optimization and other partial evolutionary algorithms. The idea of probability distribution estimation is introduced to redefine the evolutionary operation of the algorithm in discrete set space, thus improving the availability of the algorithm.
【作者单位】：华南理工大学计算机科学与工程学院;中山大学数据科学与计算机学院;
【基金】：国家自然科学基金优秀青年科学基金项目(61622206);国家自然科学基金面上项目(61379061)~~
【分类号】：TP18

【相似文献】