摘要
传统BP算法主要存在网络收敛速度慢,易陷入局部极小的问题。针对经典BP算法存在的问题,提出一种新型激励函数,并且联合使用了一些先进的技术对人工神经网络做了改进,改进算法具有更快的收敛速度、并且能有效地避免算法陷入局部极小。
the traditional BP algorithm mainly exist in the network convergence speed, easy to fall into local minimum problem. According to the classic BP algorithm problems, this paper puts forward a new incentive function, and Combined use of some advanced techniques on artificial neural network is improved, the improved algorithm has faster convergence speed, and can effectively avoid the algorithm being trapped in local minima.
出处
《科技通报》
北大核心
2012年第4期97-98,127,共3页
Bulletin of Science and Technology
关键词
神经网络
学习算法
激励函数
隐含层神经元
neural network
learning algorithm
activation function of neurons in the hidden layer