A general approach to statistical modeling of physical laws: nonparametric regression

I. Grabec

Igor Grabec update to 2007-05-23

https://arxiv.org/abs/0704.0089
Statistical modeling of experimental physical laws is based on theprobability density function of measured variables. It is expressed byexperimental data via a kernel estimator. The kernel is determined objectivelyby the scattering of data during calibration of experimental setup. A physicallaw, which relates measured variables, is optimally extracted from experimentaldata by the conditional average estimator. It is derived directly from thekernel estimator and corresponds to a general nonparametric regression. Theproposed method is demonstrated by the modeling of a return map of noisychaotic data. In this example, the nonparametric regression is used to predicta future value of chaotic time series from the present one. The mean predictorerror is used in the definition of predictor quality, while the redundancy isexpressed by the mean square distance between data points. Both statistics areused in a new definition of predictor cost function. From the minimum of thepredictor cost function, a proper number of data in the model is estimated.

journal: None

category: physics.data-an physics.gen-ph

物理定律统计建模的一般方法: 非参数回归

I. Grabec

Igor Grabec update to 2007-05-23

https://arxiv.org/abs/0704.0089
实验物理定律的统计建模基于测量变量的概率密度函数。它通过核估计器由实验数据表示。内核是由实验装置校准期间数据的散射客观确定的。通过条件平均估计器从实验数据中最佳地提取与测量变量相关的物理定律。它直接来自核估计器,对应于一般的非参数回归。通过对嘈杂混沌数据的返回图进行建模来证明所提出的方法。在此示例中,非参数回归用于从当前时间序列预测混沌时间序列的未来值。平均预测误差用于定义预测器质量,而冗余度由数据点之间的均方距离表示。这两种统计数据都用于预测成本函数的新定义。从预测器成本函数的最小值,估计模型中适当数量的数据。

期刊参考: None

category: physics.data-an physics.gen-ph