基于预测复杂性的神经网络预测子辨识  

IDENTIFICATION OF NEURAL NETWORK PREDICTOR BY MEANS OF PREDICTION COMPLEXITY

作　　者：侯越先[1] 何丕廉[1] 

机构地区：[1]天津大学计算机系,300072

出　　处：《信息与控制》2001年第1期16-20,共5页Information and Control

基　　金：国家自然科学基金 ( 6 97830 0 4)资助项目

摘　　要：本文在信息熵和互信息的基础上 ,提出了非线性偏自相关的定义 .这一概念是对线性偏自相关的一般化 ,由它可以得到度量时间序列预测复杂性的定量方法 .这种复杂性由当前序列值对各阶历史序列值不可约的依赖性所决定 ,并被非线性偏自相关的衰减趋势所反映 .通过考察这种衰减趋势 ,可以有效地进行预测模型的辨识 ,特别是神经网络这类通用非线性模型的辨识 .仿真实验很好的支持了我们的想法 .Based on information entropy and mutual information, we proposed the definition of nonlinear partial autocorrelation. The concept is the generalization of partial autocorrelation. By means of it, we could get the quantitative method to measure the intrinsic prediction complexity of time series. The complexity is determined by the irreducible dependence between current quantities of time series and high order historical quantities, and indicated by the attenuation trend of nonlinear partial autocorrelation. In according to the attenuation trend, in principle, researchers could implement nonlinear model identification, e.g., identification of neural networks. Computer simulations perfectly supported our idea.

关 键 词：预测复杂性 模型辨识 神经网络 非线性模型 时间序列 

分 类 号：O231.3[理学—运筹学与控制论] TP13[理学—数学]