一种解决群进化算法参数设置问题的最优向量法

群进化算法是智能计算领域研究的核心内容,而算法中数值型参数的设置是影响算法搜索效率的重要因素,因此设计解决参数设置问题的方法也是群进化算法研究的重要内容.目前解决参数设置问题的常规统计方法是根据算法搜索的部分结果组成有限样本数据,依据统计最好值个数大小的判定结果来确定最优参数预设值.常规统计方法在有些测试样本数据中很难确定唯一的最优参数预设值.为了解决常规统计方法的缺点,提出了一种最优向量法,该方法可以将任意形式有限样本数据转换为向量,依据向量计算的判定规则进行最优参数预设值的确定.实验结果表明,依据获取的有限样本数据通过最优向量法找到最优参数值,采用该参数值的群进化算法搜索效率相对最优,从而验证了最优向量法的有效性.

An Optimal Vector Method for Parameter Setting of Swarm Evolution Algorithm

The swarm evolution algorithm is the core content of intelligent computing research, and the setting of numerical parameters in the algorithm is an important factor affecting the search efficiency of the algorithm. Therefore, the design of the method to solve the problem of parameter setting is also an important content of swarm evolution algorithm research. At present, the conventional statistical method to solve the problem of parameter setting is to determine the optimal parameter preset value according to the result of the number of statistical optimal values based on the finite sample data composed of partial results of algorithm search. It is difficult for conventional statistical methods to determine the unique optimal parameter preset value in some test sample data. In order to solve the shortcomings of conventional statistical methods, an optimal vector method is proposed. In this method, the arbitrary finite sample data are converted into vectors, and the optimal parameters are accurately determined according to the decision rules of vector calculation. The analysis of experimental results shows that the search efficiency of the swarm evolution algorithm is relatively optimal, which uses the optimal parameter values found by the optimal vector method based on the obtained finite sample data, and the effectiveness of the optimal vector method is verified.