摘要
现有基于逻辑门的底层硬件进化方法只能进化小规模数字电路,由此提出了一种硬件系统级在线进化方法。设计了一种适合系统进化的神经元结构,采用遗传算法(SGA)来实现数字电路的在线进化设计。以图像处理算法为例,讨论分析了进化参数选取对进化结果的影响,验证了系统级进化方法的高进化成功率、快速收敛性。实验结果表明,神经元结构的设计是正确的,所提出的系统级进化方法能用于大规模数字系统的在线进化。
Since the traditional hardware evolutionary methods based on logical gates were only valid to evolve simple digital circuits, a method of on-line system-level evolution was proposed. The nerve cell model suited to digital system evolution was designed, and the Simple Genetic Algorithm （SGA） was used to realize on-line evolution of digital circuits. By example of the evolution for image processing algorithm, the influences of evolutionary algorithm parameters and template size to the evolution of image processing were analyzed. Experimental results demonstrate that the evolutionary method based on nerve cell has high evolutionary successful rate and fast convergent performance, and it can realize on-line evolution of large scale digital circuits.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2006年第5期91-94,121,共5页
Opto-Electronic Engineering
基金
国家自然科学基金(60374008)
南京航空航天大学科研创新基金(S0271-033)
关键词
进化硬件
神经元结构
系统级进化
信息处理
Evolvable hardware
Nerve cell model
System-level evolution
Information processing
