基于混合遗传算法和神经网络的软硬件划分算法  

On Hardware/Software Partitioning Algorithm Based on Genetic Algorithm and Neural Network Hybrid

作　　者：邓定胜[1] 

机构地区：[1]四川民族学院计算机科学系,四川康定626001

出　　处：《西南师范大学学报（自然科学版）》2015年第10期29-34,共6页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：四川省教育厅自然科学一般项目(15ZB0332)

摘　　要：在对比了Hopfield神经网络(Hopfield Neural Networks,HNN)和遗传算法(Genetic Algorithm,GA)各自特性的基础上,针对嵌入式系统和SoC的软/硬件双路划分问题,在时间性能的约束下,将系统功耗作为优化目标,提出了基于Hopfield神经网络和遗传算法的混合算法(GA_HNN).Hopfield神经网络快速求解的特性和遗传算法高效的全局搜索能力,使得GA_HNN算法能够跳出局部最优解而快速趋于全局最优解.最后通过实验仿真表明,GA_HNN算法与单一的Hopfield神经网络和遗传算法相比,不仅具有高速的搜索能力,而且还具有更好的节省功耗的优势.In contrast to the Hopfield neural network（Hopfield Neural Networks,HNN）and genetic algorithms（Genetic Algorithm,GA）,Hopfield neural networks and genetic algorithms hybrid algorithm（GA_HNN）have been proposed on the basis of the advantages and disadvantages of each,for embedded systems and hardware/software partitioning problem Dual SoC,under the constraints of time performance the system power consumption as the optimization objective.Because of the Hopfield neural networks and genetic algorithms feature an efficient global search capability based fast solution,making GA_HNN algorithm can jump out of local optima and quickly become a global optimal solution.At last,through the experimental simulation show that the GA_HNN algorithm compared with single Hopfield neural network and genetic algorithm,not only has high speed search ability,but also have the advantage of better save power consumption.

关 键 词：软硬件划分 Hopfiled神经网络算法 遗传算法 片上系统 嵌入式系统 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]