日长变化预报中BP神经网络拓扑结构的选择

DETERMINATION OF THE TOPOLOGY OF THE NEURAL NETWORKS IN THE PREDICTION OF LOD

日长变化的预报具有重要的科学意义和实际应用价值。非线性的人工神经网络技术中的反向传播模型(BP网络)可用于预报日长变化。BP网络的拓扑结构决定了神经网络解决问题的能力,针对不同的问题需要采用不同的网络结构。该文分析了神经网络的拓扑结构算法,选用最小均方误差法确定网络的拓扑结构,并将此应用于日长变化预报。结果表明,该方法是可靠和有效的。

The prediction of the Length of Day （LOD） is of great scientific and practical importance. This study employs the non-linear artificial neural networks （ BP network, i.e. Back-Propagation network） to predict the LOD change. The predicting ability of the BP network is determined by the topology of the network. Different topologies are needed to solve different problems. This study analyzes the algorithms of topology determinations, and chooses the least Root Mean Squared Error （ RMSE ） as a criterion to determine the topology of the network. Finally, this paper applied the developed method to predict the LOD change. The results show that this method is reliable and effective.