单指标模型异方差检验（英文）

在可加回归模型中,高维回归分析一般采用单指标模型.该模型与参数模型相比更加灵活,同时避免了维数灾难,因为单指标将标准变量向量的维数降低为单变量指标.本文构建了一个带有函数型误差项的单指数回归模型用于检验单指标模型的异方差性.由于回归模型的有效推断要求在存在异方差的情况下考虑异方差,本文提出了检验单指标模型方差不变性的假设.将Levene检验和无限因子水平的方差分析理论结合得到检验统计量用来评估方差同质性.模拟研究显示与已有方法相比,所提检验统计量适用于多种情形.最后将本文的方法应用于分析一组实际数据.

Auxiliary error and probability density function based neuro-fuzzy model and its application in batch processes

This paper focuses on resolving the identification problem of a neuro-fuzzy model(NFM) applied in batch processes. A hybrid learning algorithm is introduced to identify the proposed NFM with the idea of auxiliary error model and the identification principle based on the probability density function(PDF). The main contribution is that the NFM parameter updating approach is transformed into the shape control for the PDF of modeling error. More specifically, a virtual adaptive control system is constructed with the aid of the auxiliary error model and then the PDF shape control idea is used to tune NFM parameters so that the PDF of modeling error is controlled to follow a targeted PDF, which is in Gaussian or uniform distribution. Examples are used to validate the applicability of the proposed method and comparisons are made with the minimum mean square error based approaches.

Moderate deviations and hypothesis testing for signal detection problem

Abstract We use moderate deviations to study the signal detection problem for a diffusion model. We establish a moderate deviation principle for the log-likelihood function of the diffusion model. Then applying the moderate deviation estimates to hypothesis testing for signal detection problem we give a decision region such that its error probability of the second kind tends to zero with faster speed than the error probability of the first kind when the error probability of the first kind is approximated by e-ατ（T）, where α〉 0, τ（T） = o（T） and τ（T）→∞ as the observation time T goes to infinity.

Research on regional ionospheric tec modeling using RBF neural network

Machine learning techniques which are about the construction and study of system that can learn from data are combined with many application fields.A method on ionospheric total electron content(TEC)mapping is proposed based on radical basis function(RBF)neural network improved by Gaussian mixture model(GMM).Due to the complicated ionospheric behavior over China,GMM is used to determine the center of basis function in the unsupervised training process.Gradient descent is performed to update the weights function on a sum of squared output error function in the supervised learning process.The TEC values from the center for orbit determination in Europe(CODE)global ionospheric maps covering the period from 2007to 2010 are used to investigate the performance of the developed network model.For independent validation,the simulated TEC values at different latitudes(20°N,30°N and 40°N)along 120°E longitude are analyzed and evaluated.The results show that the simulated TEC from the RBF network based model has good agreement with the observed CODE TEC with acceptable errors.The theoretical research indicates that RBF can offer a powerful and reliable alternative to the design of ionospheric TEC forecast technologies and thus make a significant contribution to the ionospheric modeling efforts in China.

Robust estimation in inverse problems via quantile coupling

In this article we consider a sequence of hierarchical space model of inverse problems.The underlying function is estimated from indirect observations over a variety of error distributions including those that are heavy-tailed and may not even possess variances or means.The main contribution of this paper is that we establish some oracle inequalities for the inverse problems by using quantile coupling technique that gives a tight bound for the quantile coupling between an arbitrary sample p-quantile and a normal variable,and an automatic selection principle for the nonrandom filters.This leads to the data-driven choice of weights.We also give an algorithm for its implementation.The quantile coupling inequality developed in this paper is of independent interest,because it includes the median coupling inequality in literature as a special case.

ASYMPTOTIC NORMALITY OF WAVELET ESTIMATOR IN HETEROSCEDASTIC MODEL WITH α-MIXING ERRORS

Consider heteroscedastic regression model Yni= g（xni） ＋ σniεni （1 〈 i 〈 n）, where σ2ni= f（uni）, the design points （xni, uni） are known and nonrandom, g（.） and f（.） are unknown functions defined on closed interval [0, 1], and the random errors （εni, 1 ≤i≤ n） axe assumed to have the same distribution as （ξi, 1 ≤ i ≤ n）, which is a stationary and a-mixing time series with Eξi =0. Under appropriate conditions, we study asymptotic normality of wavelet estimators of g（.） and f（.）. Finite sample behavior of the estimators is investigated via simulations, too.

Comparison of Synchronization Ability of Four Types of Regular Coupled Networks

We investigate the synchronization ability of four types of regular coupled networks. By introducing the proper error variables and Lyapunov functions, we turn the stability of synchronization manifold into that of null solution of error equations, further, into the negative definiteness of some symmetric matrices, thus we get the sufficient synchronization stability conditions. To test the valid of the results, we take the Chua's circuit as an example. Although the theoretical synchronization thresholds appear to be very conservative, they provide new insights about the influence of topology and scale of networks on synchronization, and that the theoretical results and our numerical simulations are consistent.

ASYMPTOTICS FOR THE DISTRIBUTION FUNCTION ESTIMATORS OF THE ERRORS IN SEMI-PARAMETRIC REGRESSION MODELS

This paper considers the convergence rates for nonparametric estimators of the error distribution in semi-parametric regression models. By establishing some general laws of the iterated logarithm, it shows that the rates of convergence of either the empirical distribution or a smoothed version of the empirical distribution function matches exactly the rates obtained for an independent sample from the error distribution.