摘要
针对误差反向传播(BP)算法训练速度慢和易于陷入局部最小值的缺点,提出了利用遗传算法(GA)的全局寻优性,结合GA和BP的各自优点,分析和建立了进化神经网络(GA-BP)模型,并将该模型应用于似大地水准面模型精化。最后以南方某市E级GPS控制网高程数据为例,进行BP和GA-BP模型的对比实验,通过对内、外符合精度及MAPE(平均绝对误差百分比)指标分析,验证了该方法的可行性。
For error back-propagation (BP) algorithm training slowly and easily into the shortcomings of local minimum, the use of genetic algorithms (GA) global optimization is proposed. Combination of GA and BP's own merits, the evolutionary neural network (GA-BP) model is analyzed and made. This model was applied to quasi-geoid model refinement. Finally, taking E-class southern city elevation data of GPS control network for example, for BP and GA-BP experiment, through internal and external in line with the precision and MAPE analysis, the method is feasible.
出处
《全球定位系统》
2009年第4期66-70,共5页
Gnss World of China
基金
教育部博士点基金(新教师)(200802901516)
教育部博士点基金(200802900501)