摘要
提出了一种用于训练粗糙RBF神经网络(rough RBF neural networks,R-RBF)的极速学习机(extreme learning machine,ELM)方法,通过引入矩阵的Moore-Penrose逆,将传统的迭代学习方法转换为一种求线性方程的极小范数最小二乘解的方法.实验证明,在训练精度、训练时间上都能够达到非常优越的性能,其泛化精度能够提升50%以上.
The paper proposes a method of training rough RBF neural networks（R-RBF） using the extreme learning machine（ELM）, which eonverts the traditional iterative training method to solve norm least-squares solution of general linear system by introducing Moore-Penrose inverse. Experiments show that it can reach a very superior performance in both time and aeeuraey when ELM trains the Rough RBF Neural Networks, which can improve the generalization accuracy more than 50% compared with the traditional thinking of adjusting parameters iterative[y.
出处
《微电子学与计算机》
CSCD
北大核心
2012年第8期9-14,共6页
Microelectronics & Computer
基金
国家自然科学基金(41074003
60975039)
中国科学院智能信息处理重点实验室开放基金(IIP2010-1)
