局部反馈函数联接网络及其应用

Locally Recurrent Functional Link Nets and Its Applicantions

结合时间序列分析方法提出了一种具有局部反馈回路的函数联接网络（ＬＲ－ＦＬＮ）模型，并利用ＬＲＦＬＮ对非线性时间序列进行了建模与预报．对网络扩展函数的选取、初始权集的设定、以及网络结构选择和学习算法进行了研究．同时，通过对模拟数据和机床切削颤振数据的建模与预报，将其与常用的时间序列模型（ＡＲ、ＡＲＭＡ和指数自回归模型）以及ＢＰ网络进行了比较．研究结果表明该方法是可行和有效的．

A locally recurrent functional link network is proposed for modeling and prediction of nonlinear time series, which incorporates the concepts used in time series analysis. The selection of its expansion function and architecture, and its learning algorithm are also studied. With the application to simulated data and real data and the comparison to some time series models and BP network, it is shown that this method is available and more efficient than the classical models.