摘要
为了提高CMAC(cerebellar model articulation controller)神经网络实时在线学习的快速性和准确性,在基于信度分配的CA-CMAC-AMS学习算法的基础上,结合牛顿向前插公式,提出了一种新的CMAC-AMS学习算法(CA-NCMAC-AMS)。数值模拟表明,这种CA-NCMAC-AMS的学习算法不但有较快的学习速度和训练精度及建模能力,而且在信号处理、模式识别及高精度的实时智能控制等领域具有很大的应用潜力。
In order to improve the speed and accuracy of the on-line learning neural network, a new learning algorithm (CA-NCMAC-AMS) based on Credit Assignment and Newton's Forward Interpolation formula is proposed. Numerical simulations shows that the improved CMAC neural network is not only faster and more accurate than the conventional CMAC, but also has greater application potential in signal processing, pattern recognition, and controller implementation for high-precision real-time intelligent control.
出处
《北京联合大学学报》
CAS
2007年第2期1-3,共3页
Journal of Beijing Union University
关键词
CMAC神经网络
联想记忆系统
函数逼近
可信度
牛顿向前插公式
CMAC neural network
associative memory system
function approximation
reliability
Newton' s backward Interpolation formula