论文部分内容阅读
为了从传统进化策略的角度分析并改进云进化策略,研究云分布的峰度统计量及其应用.云分布在固定标准差时,也可通过调整峰度来改变噪声形状,可能产生更有效的变异.推导云分布峰度计算公式,以支持熵-超熵空间和标准差-峰度空间的相互转换.比较峰度和峰比对云分布噪声的影响,证明峰度更适宜自适应演化.给出峰度驱动的云进化策略,它的参数演化结合基于1/5规则的标准差演化和自适应峰度演化.对8个测试函数的实验结果显示,高峰度利于全局寻优,低峰度利于局部寻优,而峰度的自适应调整可综合二者优势.
In order to analyze and improve the cloud evolution strategy from the perspective of the traditional evolutionary strategy, we study the kurtosis statistics of cloud distribution and its application.When the clouds are distributed over a fixed standard deviation, the shape of the noise can also be changed by adjusting the kurtosis, which may result in more efficient The kurtosis calculation formula of cloud distribution was deduced to support the conversion between entropy-super-entropy space and standard deviation-kurtosis space.Comparing the effect of kurtosis and peak ratio on cloud distribution noise, it is proved that kurtosis is more suitable for adaptive evolution. The kurtosis-driven cloud evolution strategy is proposed, whose evolution of parameters is based on the standard deviation evolution and adaptive kurtosis evolution based on the 1/5 rule.The experimental results of eight test functions show that the kurtosis is advantageous to the global optimization and the low peak Degree conducive to local optimization, and kurtosis adaptive adjustment can be integrated both advantages.