神经网络在地震研究中的应用

APPLICATIONS OF NEURAL NETWORK TO EARTHQUAKE RESEARCHES

本文简要介绍了神经网络的模型与算法及其在地震分析、地震预报和地震工程等方面的初步应用结果。在应用于地震研究的上述三个方面时,一般都选用含两个中间层(隐层)的前向神经网络模型,并采用与之相适应的BP算法。初步研究结果表明,用神经网络方法可得出比统计学方法等更为有效的地震分析、分类、预报、震害预测等结果,因而这一方法在地震研究中有广阔的应用前景。

This paper gives a brief review of the models and algorithms of neural network (NN) and the preliminary results of its applications to seismic interpretation, earthquake prediction and earthquake engineering. In applications to the above-mentioned three aspects, a forward-type model of NN containing two hidden layers is used in most cases, and the corresponding back propagation (BP) Learning algorithm is employed usually. The preliminary results show that by using the neural network, more effective interpretation, classification, prediction and assessment in earthquake researches can be made than by using statistical and other methods. Therefore, the neural network has broad prospect in applications to the earthquake researches.