基于神经网络和遗传算法的实验参数确定

将 BP神经网络用于实验数据结果的“预处理”,剔除奇异样本数据 ,并利用实数直接操作的遗传算法 (FGA) ,发挥 FGA概率搜索和群体优势 ,将处理后的数据分别代入给定的含参数的方程 ,在总体误差最小的情况下 ,寻求符合实验数据的参数的优化解 ;将优化结果与常规数理统计方法进行了比较 ,证实了该方法的有效性 ;而且对其它实验数据的处理及实验参数的确定 ,有一定的参考价值。

Experiment Perameter Determination Based on Neural Network and Genetic Algorithms

BP neutral network was applied to preprocess the experiment data results and to eliminate the unnatural data. Using the probability search and populations characteristic of floating-point genetic algorithms (FGA),the processed data were applied to the given equation. When the total errors were minimized,the optimum parameter of the experiment data was obtained. Comparing the results with traditional mathematic statistic,it is proven that this method is more effective. At the same time,this method offers a reference value for other experiment data handled and experiment parameters attained.