可用于演化硬件的改进自适应遗传算法研究

摘要

为提高演化硬件在演化过程中的收敛速度,以解决其可扩展性问题,研究了标准遗传算法的3个遗传算子,分析了进化不同阶段对遗传算子的不同要求及其对收敛速度的影响.在Srinivas的自适应策略和基于阶段进化的自适应策略的基础上,提出一种新的针对变异算子的自适应策略,并在轮盘赌选择方式中加入适应值标度变换.结合实例,对改进后的算法进行了仿真,结果表明了加入适应值尺度变换和新的自适应策略后,算法的收敛性有所提高.%In order to improve the convergence of evolvable hardware in the process of evolution and solve the problem of scalability, three genetic operators are researched. At the same time, the different requirements to these operators in different phases and the infection on convergence rate are analyzed. On the basis of the Srinivas's adaptive strategy and phase evolution, a new adaptive strategy for mutation operator is proposed. In addition, fitness scaling is used in roulette wheel selection. The improved algorithms is emulated. It is proved that the convergence rate is indeed improved after adding fitness scaling and the new adaptive strategy.