基于分布式人工免疫算法的数值优化

A Distributed Artificial Immune Algorithm for Numerical Optimization

本文提出了一种分布式的人工免疫系统模型——塔式主从模型(TMSM),并基于此模型设计了一种用于解决数值优化问题的分布式免疫记忆克隆选择算法(DIMCSA).借助Markov模型,文中证明了DIMCSA的收敛性.为了摆脱网络连接状态对算法性能的影响,客观地衡量分布式人工免疫优化算法的性能,本文设计了多线程虚拟并行计算仿真系统,并分别考虑算法搜索时间和网络通信时间,给出了一种新的比较分布式随机搜索算法性能的指标.实验结果表明,DIMCSA能够用较少的计算代价和通信代价获得更高质量的解,适合解决大规模的复杂优化问题.

This paper proposes a distributed model termed as Tower-like Master-Slave Model （TMSM） for the artificial immune systems.Based on TMSM, a distributed immune memory clonal selection algorithm （DIMCSA） is put forward for solving numerical optimization problem. Using the theorem of Markov chain, we have proved the convergence of DIMCSA. In order to get away from the influence of network conditions and get a veracious estimation on the DIMCSA＇ efficiency,Multi-thread simulative parallel computing system （MSPCS） is designed here and a novel performing index in which the searching time and network communication time are considered respectively is also proposed for distributed stochastic searching approaches. Experimental results indicate that DIMCSA can achieve better solutions with less computing and fewer communications, and it is capable of solving massive and complicated optimization problems.