摘要
泛化能力是人工神经网络的重要特性.总结了增强神经网络泛化能力的几种方法,分析了各自的优缺点,并使用Matlab中提供的函数进行非线性函数的逼近,通过仿真与原函数拟合验证了泛化能力的提高.
The generalization capability is one most important performance of Artificial Neural Network （ANN）. This paper discusses several methods on enhancement of ANN generalization capability, and analyzes their advantages and disadvantages. Functions provided by Matlab were employed to approximate the non-linear function, simulation and fitting with original fuction were carried out to verify the improvement of generalization capability.
出处
《南京信息工程大学学报（自然科学版）》
CAS
2011年第2期164-167,共4页
Journal of Nanjing University of Information Science & Technology（Natural Science Edition）
基金
国家重点基础研究发展计划项目(2009CB320501)
关键词
神经网络
泛化能力
MATLAB
artificial neural networks
generalization capability
Matlab
