摘要
免疫算法借鉴了生物免疫系统独有的自适应、自组织、多样性、免疫记忆等优良特性,是智能计算领域中继人工神经网络和进化计算之后的又一个研究热点.提出一种新型的基于聚类的免疫多目标优化算法(CMOIA),借鉴了免疫算法的亲和度定义,由此亲和度定义的免疫变异操作子使得算法产生的抗体群体能够逐渐向精英群体变异,结合进化算法在局部搜索中维持解个体多样性的能力对免疫变异产生的抗体群进行交叉变异操作,采用一种基于聚类的克隆选择算子来保持免疫算法在探测新解和加强局部搜索之间的平衡.选取了8个通用的多目标优化问题对3个广泛采用的性能指标进行了测试.与现有两个经典的进化优化算法相比较,算法所产生的解集在收敛性、多样性等方面显示了相当的独创性和先进性.
Immune algorithms have extracted from the characteristics of biological immune system, such as self-adaptation, self-organ- ization, diversity, immune memory, etc. It was another hotspot after artificial neural network and evolutionary computation. In the paper, a novel Clustering based Multi-objective Optimization Immune Algorithm （CMOIA） is proposed. It refers to affinity defini- tion of the immune algorithm, the mutation operator based on this definition can make the generated antibodies develop into a much better group. It combines local search ability of evolutionary algorithm by using crossover and genetic mutation operators to operate on the immune mutated antibodies. Then using a clustering based clonal selection operator to maintain a balance between exploration and exploitation. In addition, eight general multi-objective optimization problems are selected to test according the widely used three per- formance indicators. It was shown that the Pareto Fronts obtained by CMOIA were better convergence and diversity than the ones from the other existing two classical multi-objective optimization evolutionary algorithms.
出处
《小型微型计算机系统》
CSCD
北大核心
2012年第9期1948-1953,共6页
Journal of Chinese Computer Systems
基金
国家博士后基金项目(20090451241)资助
关键词
人工免疫
多目标优化
克隆选择
聚类
artificial immune systems
multi-objective optimization
clonal selection
clustering
