摘要
为克服传统的BP网络的不足 ,采用自适应变步长算法 (ABPM)来训练前馈人工神经网络。根据黄河流域的大汶河水系的水质监测的数据 ,建立了一个对地面水质进行判别的多层前馈网络数学模型。以地面水质七项污染指标为训练样本 ,对网络进行训练 ,并将训练好的网络用于水质进行评价 ,将计算结果与BP网络评价结果、单因子评价结果进行了比较分析。结果表明 ,ABPM神经网络方法收敛速度较快 ,预测精度很高 。
The back-propagation (BP) network is trained using adaptive variable step size algorithm to overcome the shortage of conventional BP Network.Based on the observation data the ANN model for water quality evaluation is established.Seven pollution indices are adopted as training stylebook to train the network.The trained network can be used to evaluate the water quality.The contrast comparison between the conventional BP network and proposed network shows that the later possesses the advantage of high accuracy and high speed for convergence.
出处
《水利学报》
EI
CSCD
北大核心
2002年第10期119-123,共5页
Journal of Hydraulic Engineering
