摘要
对Ｌｏｇｉｔ回归和前馈神经网络作了一个尝试性的“比较研究”，说明了一些理论结果和特性，讨论了应用中的一些实际问题。并进一步用分析和模拟两种方法研究了渐近概念、过分拟合以及模型选择等问题。
Attempt is made to undertake comparative study of feedforward neural network and Logit regression model, whereby some theoretical results and properties are clarified with their application to practical problems. Also, some aspects of approximation, overfitting and model selection are examined analytically and numerically.
出处
《南京气象学院学报》
CSCD
1997年第4期468-478,共11页
Journal of Nanjing Institute of Meteorology
关键词
前馈神经网络
LOGIT回归模型
模型选择
feedforward neural networks, Logit regression model, NewtonRaphson algorithm, model selection
