多维非线性自回归模型的投影寻踪学习网络逼近

本文研究基于投影寻踪学习网络的多维非线性自回归模型逼近的收敛性，证明了在L^k(k为正整数）空间上，投影寻踪学习网络可以任意精度逼近多维非线性自回归模型，并给出应用实例。     

非线性系统高维特征量的稳健投影寻踪建模

针对非线性系统高维特征量的识别与提取问题,本文给出了稳健投影寻踪建模的方法。应用此方法,对试飞实测数据进行处理,建立了飞机发动机低压转子转速与其余六个特征量的稳健投影寻踪模型。上述模型不仅揭示了该非线性系统多个特征量之间关系,而且模型精度高,达到了实际工程要求。     

投影寻踪模型的改进及其在城市水资源承载能力预测中的应用

在城市水资源承载能力研究中,偏最小二乘回归方法能有效地处理自变量间多重线性相关性问题,但不能较好地处理因变量与自变量间复杂的非线性问题.投影寻踪神经网络耦合模型是处理非线性问题的有力工具,而且神经网络投影寻踪耦合模型稳健性高,但不能较好地处理自变量间多重线性相关性问题.本文把这两种方法结合在一起,建立了基于偏最小二乘回归的神经网络投影寻踪耦合模型,对城市水资源承载能力进行了预测,并取得了满意效果.     

基于ARIMA和LSSVM的非线性集成预测模型

针对复杂时间序列预测困难的问题,在综合考虑线性与非线性复合特征的基础上,提出一种基于ARIMA和最小二乘支持向量机(LSSVM)的非线性集成预测方法.首先采用ARIMA模型进行时间序列线性趋势建模,并为LSSVM建模确定输入阶数;接着根据确定的输入阶数进行时间序列样本重构,采用LSSVM模型进行时间序列非线性特征建模;最后采用基于LSSVM的非线性集成技术形成一个综合的预测结果.将该方法用于中国GDP预测取得的结果,与单独预测方法及流行的其他集成预测方法相比,预测精度有了较大的提高,从而验证了方法的有效性和可行性.     

基于SVR的RAR区域经济预测模型

由于区域经济系统中许多经济变量呈现出强非线性与大波动性的特征,使得传统的时间序列线性建模和预测技术难以适应区域经济预测的要求.为此,提出基于支持向量机改进的残差自回归区域经济预测模型.首先采用时间序列分析中的残差自回归模型对时间序列趋势进行线性拟合,然后对残差自回归模型估计后的残差序列采用支持向量回归方法再次提取其非线性特征,从而提高区域经济时间序列模型的预测精度.最后以广东省GDP的预测实例说明模型的有效性.     

回归函数的投影寻踪逼近的Lp收敛性

投影寻踪是用来处理高维数据得一类新型统计方法。由于不知道Ｅ（ｒｍ（ｘ）│α＾Ｔｍｘ）的具体形式,给投影寻踪回归的应用带来一定的困难,为此,作者曾证明岭函数为多项式形式的投影寻踪回归的Ｌ２收敛性「３」,本文在文献「３」的基础上进一步证明了回归函数投影寻踪逼近的Ｌｐ收敛性。     

回归函数的投影寻踪逼近的L_p收敛性

投影寻踪是用来处理高维数据得一类新型统计方法．由于不知道Ｅ（ｒｍ（ｘ）｜ａＴｍｘ）的具体形式，给投影寻踪回归的应用带来一定的困难，为此，作者曾证明岭函数为多项式形式的投影寻踪回归的 Ｌ２收敛性［３］，本文在文献 ［３］的基础上进一步证明了回归函数投影寻踪逼近的 Ｌｐ收敛性．     

基于稀疏自编码器的金融市场指数预测模型

针对高维数据"维数灾难"问题,降维是最典型的处理方式之一。降维技术不仅可以减弱"维数灾难"的负面影响,而且能够剔除高维数据中的冗余特征,从而提升高维数据回归、分类等任务的效率。高维数据通常呈现出复杂或非线性结构,恰当的降维方法可以有效地将高维特征数据投影至低维空间,以实现原始数据的非线性特征提取。本文尝试使用无监督学习模型稀疏自编码网络对金融高维数据进行非线性特征提取,将提取到的特征作为有监督学习模型BP神经网络的输入以预测指数收益率。更进一步地,为了验证稀疏自编码算法在特征提取方面的优势与有效性,本文引入稀疏主成分模型进行对比分析。实证分析显示:本文所使用的稀疏自编码网络能够较好地提取非线性特征并进行预测,其预测精度优于以稀疏主成分为代表的线性降维方法。     

Autoregression Model of Time Series with Matrix Cross-Section Data北大核心CSCD

矩阵型截面数据时间序列的优点在于可以同时刻画多个对象的多个属性.本文重点研究了矩阵型截面数据时间序列的自回归模型,给出了该模型的参数估计、模型识别、白噪声检验三个方面的理论结果.最后再利用矩阵型截面数据时间序列自回归模型,对两支银行股的日收益率序列和日成交量变化率序列进行建模分析.     

太阳辐射时间序列的非线性检验

检验太阳辐射时间序列是否有非线性特征,对于分析、建模和预测太阳辐射量是重要、有益的.提出用基于替代数据的检验方法来检验太阳辐射时间序列是否存在非线性特征,并将数据序列的三阶矩作为检验统计量.选取了美国Montana州Dillon地区和Wyoming州Green Rivet地区每日总辐射量、Utah州Moab地区的每月日平均总辐射量时间序列作为检验对象.数值分析的统计结果表明所研究的日总辐射时间序列存在非线性,而每月日平均总辐射时间序列未检测出非线性.因而,对太阳辐射时间序列建模和预测之前,检验其是否有非线性特征是必要的.     

B值移动平均过程的Strassen重对数律

设 {ε,εt;t∈ Z}是 iid的 B值随机变量序列 ,{ aj;j∈ Z}是一个实数列 ,满足 ∞j=-∞|aj|<∞ .记 Xt= ∞j=-∞ajεt-j,Sn = nt=1Xt.对 p≥ 1 ,本文研究了n-1 -( p/ 2 ) (2 L2 n) -( p/ 2 ) ni=1 ‖ Si‖p 及 n-1 -( p/ 2 ) (2 L2 n) -( p/ 2 ) ni=0 ‖ Sn- Si‖ p的渐进性质 ,使得 Strassen(1 964)及 Chen(1 994)的一些结果得到推广 .     

A computer realization of control-with-guide procedures

Positional differential games of pursuit with target for-conflict-control systems, non-linear with respect to the phase vector, are considered. The problems investigated are approximate construction of the set of positional absorption and construction of control procedures guaranteeing guidance to the target. Issues relating to the development of algorithms for the approximate construction of the positional absorption set and control-with-guide procedures [1–4] are also examined. As an example to illustrate the possibilities of the algorithms considered, a pursuit game of the Homicidal Chauffeur type [5] is considered, with approximate computation of positional absorption sets in the problem of pursuit over a fixed time interval for several parameter sets. Motions of a conflict-control system that generate a control-with-guide procedure are computed for several specific initial values of the phase vector and several choices of the evader's control.     

Grand Tour and Projection Pursuit

Abstract

The grand tour and projection pursuit are two methods for exploring multivariate data. We show how to combine them into a dynamic graphical tool for exploratory data analysis, called a projection pursuit guided tour. This tool assists in clustering data when clusters are oddly shaped and in finding general low-dimensional structure in high-dimensional, and in particular, sparse data. An example shows that the method, which is projection-based, can be quite powerful in situations that may cause grief for methods based on kernel smoothing. The projection pursuit guided tour is also useful for comparing and developing projection pursuit indexes and illustrating some types of asymptotic results.     

Recurrent Extensions of Real-Valued Self-Similar Markov Processes

Potential Analysis - Let X = (Xt,t ≥?0) be a self-similar Markov process taking values in $\mathbb {R}$ such that the state 0 is a trap. In this paper, we present a necessary and...     

一类超一致椭圆扩散过程的极限定理

本文研究了分支特征为ψ（x,z）=γz^1+β（0<β≤1）形式的超一致椭圆扩散过程,当初始值X₀（dx）为底过程的某类不变测度时,给出了当空间维数d满足βd≤2时,超过程X_t依分布收敛于0测度,当βd>2时,X_t则依分布收敛于一个非退化的随机测度.     

Exponential change of measure for general piecewise deterministic Markov processes

We consider a general piecewise deterministic Markov process(PDMP) X = {X_t}_(t≥0) with a measure-valued generator A, for which the conditional distribution function of the inter-occurrence time is not necessarily absolutely continuous. A general form of the exponential martingales that are associated with X is given by■By considering this exponential martingale to be a likelihood-ratio process, we define a new probability measure and show that the process X is still a general PDMP under the new probability measure. We additionally find the new measure-valued generator and its domain. To illustrate our results, we investigate the continuous-time compound binomial model.     

Sliding mode control of the forced generalized Burgers equation

^** Email: smaoui{at}mcs.sci.kuniv.edu.kw This paper deals with the sliding mode control (SMC) of theforced generalized Burgers equation via the Karhunen-Loève(K-L) Galerkin method. The decomposition procedure of the K-Lmethod is presented to illustrate the use of this method inanalysing the numerical simulations data which represent thesolutions of the forced generalized Burgers equation for viscosityranging from 1 to 100. The K-L Galerkin projection is used asa model reduction technique for non-linear systems to derivea system of ordinary differential equations (ODEs) that mimicsthe dynamics of the forced generalized Burgers equation. Thedata coefficients derived from the ODE system are then usedto approximate the solutions of the forced Burgers equation.Finally, static and dynamic SMC schemes with the objective ofenhancing the stability of the forced generalized Burgers equationare proposed. Simulations of the controlled system are givento illustrate the developed theory.     

Supercritical superprocesses: Proper normalization and non-degenerate strong limit

Suppose that X = {X_t, t≥0; P_μ} is a supercritical superprocess in a locally compact separable metric space E. Let φ0 be a positive eigenfunction corresponding to the first eigenvalue λ_0 of the generator of the mean semigroup of X. Then Mt := e~(-λ_0t)〈φ0,X_t〉 is a positive martingale. Let M_∞ be the limit of M_t. It is known(see Liu et al.(2009)) that M_∞ is non-degenerate if and only if the L log L condition is satisfied. In this paper we are mainly interested in the case when the L log L condition is not satisfied. We prove that, under some conditions, there exist a positive function γ_t on [0,∞) and a non-degenerate random variable W such that for any finite nonzero Borel measure μ on E,lim/t→∞γ_t〈φ0,X_t〉=W, a.s.-P_μ~.We also give the almost sure limit of γ_t〈f, X_t〉for a class of general test functions f.     

基于投影寻踪和最优分割的企业信用评级模型

提出一种基于投影寻踪和最优分割的企业信用评级模型。该模型运用投影寻踪对样本企业进行信用综合评分,将信用综合得分由大到小排序,生成有序样品序列;利用最优分割法对有序样品进行聚类,得出明确的聚类结果;将最优分割点对应的信用综合得分作为划分信用等级的阈值,从而实现对样本企业的信用评级。应用实例证明了该模型的可行性和有效性。     

基于迭代投影的梯度硬阈值追踪算法

梯度硬阈值追踪算法是求解稀疏优化问题的有效算法之一.考虑到算法中投影对最优解的影响,提出一种比贪婪策略更好的投影算法是很有必要的.针对一般的稀疏约束优化问题,利用整数规划提出一种迭代投影策略,将梯度投影算法中的投影作为一个子问题求解.通过迭代求解该子问题得到投影的指标集,并以此继续求解原问题,以提高梯度硬阈值追踪算法的计算效果.证明了算法的收敛性,并通过数值实例验证了算法的有效性.