一种实现全父辈交叉遗传算法的神经计算模型  

A Neural Computing Model for Entire Parens Recombination Genetic Algorithm

作　　者：龚道雄[1] 阮晓钢[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100022

出　　处：《计算机工程与应用》2004年第15期13-16,174,共5页Computer Engineering and Applications

基　　金：国家自然科学基金重点项目资助(编号:60234020)

摘　　要：该文提出了一种实现二值编码全父辈交叉遗传算法的神经计算模型GAnn。GAnn将遗传算法的迭代循环结构通过一组由神经元模块所组成的反馈回路实现,并给出了实现二值编码全父辈交叉操作以及突变操作的人工神经元和神经网络拓扑结构。该文通过实验结果验证了GAnn的可行性。GAnn综合了硬件遗传算法和并行遗传算法的优点,这对于用硬件实现遗传算法,显式地实现遗传算法的内在并行性,提高遗传算法的实时性,拓宽遗传算法应用领域的研究具有积极的意义。This paper proposes a neural computing model for the binary coded Entire Parents Recombination Genetic Algorithm for the sake of integrating the hardware genetic algorithm and the parallel genetic algorithm.In the model,the evolving iteration of GA is implemented with a feedback loop composed of neural network modules.The neural networks as well as the neurons of both binary coded entire parents recombination and mutation operation are designed.Simulations on a suit of benchmark functions validate the model.This research may lead a new way of improving the real-time performance of GA,which not only implement GA with hardware but also realize the inherent parallelism of GA explicitly.

关 键 词：遗传算法 神经网络 优化计算 并行遗传算法 硬件遗传算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]