摘要
针对标准B-P神经网络算法存在学习速度慢的问题提出了改进算法。将改进后的神经网络模型应用于直接顶分类,不论是直接顶初次跨落步距的拟合值还是其预测值,神经网络法的计算精度均高于多元线性回归法。
The widely used conventional B-P neural network has been improved to speed up the training process. The new proposed neural network is used to study the classification problem of immediate roof. Be it the approximated caving step or the predicted caving step, the proposed neural network is better in accuracy than the multi-variate linear regression method.
出处
《黑龙江科技学院学报》
1994年第2期27-32,共6页
Journal of Heilongjiang Institute of Science and Technology
关键词
直接顶分类
人工神经网络
B-P学习算法
非线性建模
classification of immediate roof, artifical neural network, B-P learning algorithm, non-linear modelling
