摘要
借鉴计算机网络拥塞控制中的"慢启动"策略,针对传统BP算法中存在的收敛速度慢与精度不高的不足提出了两种改进的变学习率学习算法,仿真结果表明改进的BP算法与自适应附加动量BP算法性能相近,其学习的收敛速度与精度优于传统的BP算法.
To improve the convergence speed and learning precision,two kinds of improved Bp learning algorithms are given by introducing the `slowstart' strategy for congestion control in networks.The simulation results show that the performance of the presented algorithms is similar to the adaptive learning algorithm with momention, but better than the traditional Bp algorithms ,both in convergence speed and learning precision.
出处
《长沙大学学报》
2004年第4期54-57,共4页
Journal of Changsha University
关键词
神经网络
BP算法
变学习率
网络拥塞控制
neural network
BP algorithm
learning rate
congestion control in network
