The mutation operator has been seldom improved because researchers hardly suspect its ability to prevent genetic algorithm (GA) from converging prematurely. Due to its importance to GA, the authors of this paper study...The mutation operator has been seldom improved because researchers hardly suspect its ability to prevent genetic algorithm (GA) from converging prematurely. Due to its importance to GA, the authors of this paper study its influence on the diversity of genes in the same locus, and point out that traditional mutation, to some extent, can result in premature convergence of genes (PCG) in the same locus. The above drawback of the traditional mutation operator causes the loss of critical alleles. Inspired by digital technique, we introduce two kinds of boolean operation into GA to develop a novel mutation operator and discuss its contribution to preventing the loss of critical alleles. The experimental results of function optimization show that the improved mutation operator can effectively prevent premature convergence, and can provide a wide selection range of control parameters for GA.展开更多
There has been a growing interest in mathematical models to character the evolutionary algorithms. The best-known one of such models is the axiomatic model colled the abstract evolutionary algorithm. In this paper, we...There has been a growing interest in mathematical models to character the evolutionary algorithms. The best-known one of such models is the axiomatic model colled the abstract evolutionary algorithm. In this paper, we first introduce the definitions of the abhstract selection and evolution operators, and that of the abstract evolutionary algorithm, which describes the evolution as an abstract stochastic process composed of these two fundamental abstract operators. In particular, a kind of abstract evolutionary algorithms based on a special selection mechansim is discussed. According to the sorting for the state space, the properties of the single step transition matrix for the algorithm are anaylzed. In the end, we prove that the limit probability distribution of the Markov chains exists. The present work provides a big step toward the establishment of a unified theory of evolutionary computation.展开更多
基金National Natural Science Foundation of China(No.61806006)A Project Funded by the Priority Academic Program Development of Jiangsu Higher Education InstitutionsSupported by the 111 Project(No.B12018)。
文摘在求解多目标优化问题上,个体的多样性影响了求得解集的质量,决定了解集的分布性。为了扩大个体的搜索方向,增大种群的多样性,同时避免个体在变异过程中聚集在边界处,提出了一种具有反向混沌映射的平均突变修补差分进化(Opposition chaotic initialization and average mutation repair-based differential evolution,OCI-AMR_DE)算法。首先,为了生成均匀分布的初始种群,在初始化过程中对随机数进行tent混沌映射和反向学习,以生成均匀分布的随机数。其次,对突变算子进行处理以避免个体聚集在边界,从而提高种群的多样性。在每次迭代中,对产生突变的个体进行平均突变修补,做合法化处理,计算分别基于Pareto优势和约束优势原则(Constrained dominance principle,CDP)排名的加权和。然后,根据加权和排序选择前N个个体进入下一代,重复上述过程至满足结束条件得到结果集。最后,选取3个测试函数集共38个多目标优化问题对所提算法的性能做出评估,并将其与7种算法进行对比。实验结果表明,OCIAMR_DE在解决受约束多目标优化问题方面具有较强的竞争力。
文摘The mutation operator has been seldom improved because researchers hardly suspect its ability to prevent genetic algorithm (GA) from converging prematurely. Due to its importance to GA, the authors of this paper study its influence on the diversity of genes in the same locus, and point out that traditional mutation, to some extent, can result in premature convergence of genes (PCG) in the same locus. The above drawback of the traditional mutation operator causes the loss of critical alleles. Inspired by digital technique, we introduce two kinds of boolean operation into GA to develop a novel mutation operator and discuss its contribution to preventing the loss of critical alleles. The experimental results of function optimization show that the improved mutation operator can effectively prevent premature convergence, and can provide a wide selection range of control parameters for GA.
基金Supported by the National Science Foundation of China(60133010)Supported by the Science Foundation of Henan Province(2000110019)
文摘There has been a growing interest in mathematical models to character the evolutionary algorithms. The best-known one of such models is the axiomatic model colled the abstract evolutionary algorithm. In this paper, we first introduce the definitions of the abhstract selection and evolution operators, and that of the abstract evolutionary algorithm, which describes the evolution as an abstract stochastic process composed of these two fundamental abstract operators. In particular, a kind of abstract evolutionary algorithms based on a special selection mechansim is discussed. According to the sorting for the state space, the properties of the single step transition matrix for the algorithm are anaylzed. In the end, we prove that the limit probability distribution of the Markov chains exists. The present work provides a big step toward the establishment of a unified theory of evolutionary computation.
