摘要
神经网络被引入GPS向量网的平差处理时,网络训练数据由于在测量时各因数的干扰多少都有噪声或异常值掺入。为了使网络具有更好抵抗粗差的能力,文中根据稳健统计有关原理,针对神经网络论证了一种较好的抗差估计函数作为网络目标函数。以函数逼近来说明平差所采用的模型与神经网络处理时的异同,理论分析和实验证明了神经网络的抗粗差能力。
Introduced Neural Network to the survery of GPS net, the training date of network will be interpolated noise more or less. In order to make the network resist the noise better. According to the principle of robust statistics,this paper will propose a better analysis function to be the target. Taking the function approach for example,it explains the difference between the model of GPS net and the Neural Network with using the same date. The theoretical analysis and the experiment will proof the ability of this method ： Both of them will be manifested the most superior characteristic.
出处
《城市勘测》
2008年第3期68-71,共4页
Urban Geotechnical Investigation & Surveying
