摘要
BP神经网络被广泛应用于分类模式识别、图像处理和系统控制等领域人们对BP网络算法进行了许多的研究,但尚有其不足之处为完成其权的训练,问题的关键在于如何避免陷入局部极小及在此前题下如何提高学习速度为此,就如何选取学习率η和动量矩α提出了改进方案,并应用于数字识别。
BP neural networks have been widely used in mode recognizing, image processing, system control etc A lot of study has been done on the training algorithm of BP neural networks But there are still many short comings To complete the training of the weights of a BP neural network, the key problem is how to overcome the staying at the local minimum points as well as how to increase the training speed based on this condition This paper comes up with an improvement plan as to the selection of the learning rate η and the momentum α When this plan is applied to digital recognizing,a quite satisfactory result is achieved.$$$$
出处
《江苏理工大学学报（自然科学版）》
2000年第1期79-82,共4页
Journal of Jiangsu University of Science and Technology(Natural Science)
