基于多目标并行基因表达式编程的电路演化算法

为提高数字电路演化的效率和成功率,在并行基因表达式编程的基础上,对电路设计中涉及的多个目标进行了定义与量化,并针对这些目标提出基于多目标并行基因表达式编程的电路演化算法(MPGEP).主要工作包括:1)设计演化电路中的GEP编码;2)利用OpenMP设计基于通用多核处理器的并行基因表达式编程模型;3)定义和量化电路演化的多个目标,利用非支配排序和适应度共享策略来提高搜索方向的空间均匀性;4)通过数字电路演化实验证明,与传统的GP和GEP算法相比,MPGEP算法不仅进化时间减少了86.1%和31.4%,同时还能得到更简单和实用的电路,得到最优电路比率提高了50.4%和38.9%;与多目标串行电路演化算法MGEP相比,MPGEP算法的进化时间减少了48.7%;与并行电路演化算法PGEP-MC相比,MPGEP算法得到最优电路的比率提高了38.3%.

Evolutionary algorithm of circuits based on multiobjective parallel gene expression programming

To improve the evolution efficiency and success rate of digital circuits,the multiobjectives during evolution of circuit are defined and quantified.An evolutionary algorithm of circuits based on multiobjective parallel gene expression programming(MPGEP) is presented.The main contributions include:1) the chromosome encoding of GEP in evolution of circuits is designed;2) the parallel model of GEP based on general multi-core processors is designed by OpenMP;3) the multiobjectives during evolution of circuits are defined and quantified.And the uniformly scattered search direction is enhanced by non-dominated sorting and fitness sharing strategy;4) the experiments on evolution of digital circuits show that MPGEP improves the evolutionary efficiency.Compared with GP and GEP,the evolutionary time of the MPGEP drops 86.1% and 31.4%.MPGEP is also capable of searching out simple and practical circuits.The ratio of searching optimal circuits increases 50.4% and 38.9%.Compared with MGEP,the evolutionary time of MPGEP drops 48.7%.Compared with PGEP-MC,the ratio of searching optimal circuits of MPGEP increases 38.3%.