下标降序法最近邻判别分析

§1 引言设(X_1,θ_1),…,(X_n,θ_n),(X,θ)为在R~d×{1,…,M}上取值的i.i.d.随机向量。问题是要利用X的观察值及历史样本(X_i,θ_i),i=1,…,n对类别变量θ进行判别。假定在R~d上给定了某一距离函数p(·,·)(比如欧氏距离等),那么可按照诸X_i与X的距离由小到大把诸X_i重新排列为X_(R1),X_(R2),…,X_(R_n,相应的θ_i也被排列为θ_(R1),θ_(R2),…,θ_(Rn)。若采用θ_(R1)来判别θ,这就是所谓的最近邻判别法。Derroye,Wagner,Fritz,陈希孺及白志     

ρ-混合序列部分和乘积的几乎处处极限定理

设{X_n,n≥1}是一严平稳的ρ-混合的正的随机变量序列,且EX_1=μ＞0, Var(X_1)=σ~2,记S_n=Σ_(i=1)~n X_i和γ=σ/μ,在较弱的条件下,证明了对任意的x,,其中σ_1~2=1+2/(σ~2)∑_(j=2)~∞Cov(X_1,X_j),F(·)是随机变量e~(2~(1/2)N)的分布函数,N是标准正态随机变量,我们的结果推广了i．i．d时的情形．     

方差估计的随机加权逼近及随机加权法在抽样调查中的初步应用

设X_1,X_2,…是一组独立同分布的随机变量序列,其方差μ_2是待估参数,当x_4,i=1,2,…,n,给定下,用D_n=sum from i=1 to n(V_(ni)(X_i-sum from i=1 to n(V_(ni)X_i)~2)-1/n sum from i=1 to n(X_i-X)~2的条件分布来渐近T_n=(1/n)sum from i=1 to n(X_i-X)~2-μ_2的分布。这里D_n中的V_(ni),i=1,2,…,n,是服从 Dirichlet分布D(4,4,…,4)的随机变量。若记 F_n和F_n~*分别是T_n/(VarT_n)~(1/2)的分布和D_n/(Var~*D_n)~(1/2)的条件分布,其中Var~*D_n是关于X_1,X_2,…的条件方差。则在一定条件下,对几乎所有的样本序列X_1,X_2,…, (i)n~(1/2)D_n→N(0,μ_4-μ_2~2) 其中μ_4=E(X_1-μ)~4,μ=EX_1 ii)n(1/2)sup|F_n~*(y)-F_n(y)|=0(1) iii) lim sup |F_n~*(y)-F_n(y)|=0 最后,本文对随机加权法如何应用于抽样调查之中,进行了一个初步的尝试。     

EXPONENTIAL BOUNDS OF POSTERIOR RISK FOR k-NN PREDICTION

Let (X,θ),(X_1,θ_1),…,(X_n,θ_n) be R~d×R~1-valued random vectors, it is desired to predict the value of θ, based on the observed value of X and with the help of the training sample Z~n={(X_i,θ_i), i=1,…,n}. Cover(1) used the k-Nearest Neighbor method to this problem, the method is as follows: Introduce a metric ‖X-Y‖ in R_d. Rearrange X_1,…,X_n into X_(n1),…, X_(nn), such that ‖X_(n1)-X‖≤ ‖X_(n2)-X‖≤…≤‖X_(nn)-X‖,and break ties by comparing indices. Choose positive integer k. Denote by θ_(ni) the θ-value associated with X_(ni), i. e., θ_(ni) =θ_j when X_(ni)=X_j. Under the square loss L(θ,α)=     

强相依高斯序列的最大值与和的联合渐近分布

{X_n,n≥1)是标准高斯序列,T_(ij)=cov(X_i,X_j)。本文在强相依条件rijlog(j-i)→r∈(0,∞)(j-i→ ∞)下,得到了高斯序列的最大值M_n与标准化部分和S_n=sum from i=1 to n(X_i/(E(sum from i=1 to n X_i)~2)/(1/2))     

On necessary and sufficient conditions for the self-normalized central limit theorem In Honor of Professor Chuanrong Lu on His 85th Birthday

Let X_1, X_2,... be a sequence of independent random variables and S_n=sum X_1 from i=1 to n and V_n~2=sum X_1~2 from i=1 to n . When the elements of the sequence are i.i.d., it is known that the self-normalized sum S_n/V_n converges to a standard normal distribution if and only if max1≤i≤n|X_i|/V_n → 0 in probability and the mean of X_1 is zero. In this paper, sufficient conditions for the self-normalized central limit theorem are obtained for general independent random variables. It is also shown that if max1≤i≤n|X_i|/V_n → 0 in probability, then these sufficient conditions are necessary.     

The Berry-Esseen Theorem and Non-Uniform Estimate for U-Statistics

Let X_1,X_2,…,X_n be independent random variables. Define a U-statistic by U_n(?)~(-1)sum from 1≤i≤j≤n (h(X_i,X_j), where h(x,y) is a symmetric function of two variables x,y and that Eh(X_i,X_j)=0(i≠j, i,j=1,2,…,n). Write g_j(X_i)=E(h(x_i,x_j)|x_i),g(X_1)=1/n-1 sum from j=1 j≠i to n g_j(X_i) We give the following two theorem: Theorem 1 Suppore that     

强混合平稳序列的随机容量,随机秩顺序统计量的极限分布

§1.引言 设{X_n,n≥1}为严平稳随机序列,一维边际分布函数为F(x),W_1(n)≤W_2~(n)≤…≤W_n~(n)为X_1,X_2,…,X_n的顺序统计量.若K_n≤n为正整数列,则称W_K_n~(n)为非随机秩顺序统计量,秩为K_n;若M_n,N_n,n≥1为两正整值随机序列,N_n≤M_n,a.e.,     

一种方式分组随机模型中估计方差分量的最小风险设计

§1.引言一种方式分组随机模型:y_(ij)=β α_i ε_(ij),i=1,…,n,j=1,…,m_i,(1.1)其中 ε_(ij)(i=1,…,n,j=1,…,m_i)是相互独立的随机误差,α_i(i=1,…,n)是独立的随机变量.Eα_i=Eε_(ij)=0,varε_(ij)=θ_1>0,varα_i=θ_2≥0,cov(α_i,ε_(ij))=0.β、θ_1、θ_2是未知参数,β∈R~1,(θ_1,θ_2)~T∈Θ(?){θ_1>0,θ_2≥0}.     

一般损失下k近预测条件风险的指数界

<正> 设(X,θ)为R~d×R~1上随机变量.(X_1,θ_1),…,(X_n,θ_n)为它的独立同分布样本.设X的值已观测,记Z_n=((X_1,θ_1),…,(X_n,θ_n)),要用X和Z_n的值去预测θ的值.设‖·‖为R~d中欧氏距离或最大分量模距离,将X_1,…,X_n重排为X_(n1),…,X_(nn).使得‖X_(n1)‖-X‖≤‖X_(n2)-X‖≤…≤‖X_(nn)-X‖,以θ_(n1),…,θ_(nn)记相应的匹     

一类特殊分布纪录值之和的中心极限定理

摘要：设{X，Xn,n≥1)为独立同分布的服从某连续分布F的随机变量序列，X^(1)=X1，X^(2),X^(3),…为其纪录值序列．令ψ(u)=F^-1(1-e^-u)．其中F^-1是F的反函数．本文研究当ψ(u)=log^pu时Tn=∑k=1^nX^(k)=^dn∑k=1^nψ(Sn)的极限性质．解决了户为所有正整数时Tn的中心极限定理．     

COMPARISON OF PRODUCTS OF THE VARIETIES OF LIE ALGEBRAS

Abstract

In Fortes (2001), we introduced a notion of order for associative pairs and we obtained a Goldie-like characterization of left orders in a semiprime pair coinciding with its socle. In this paper, we take up again that notion of order to establish a Faith-Utumi theorem, which studies left orders in a prime pair coinciding with its socle.     

ADDITIVE FAMILIES OF INVARIANTS OF FORMS OF DEGREE d

Abstract

Let d be a positive integer, and F be a field of characteristic zero. Suppose that for each positive integer n, I _n, is a GL _n,(F)- invariant of forms of degree d in x₁, …, x _n, over F. We call {I _n} an additive family of invariants if I _p+q (f ⊥ g) = I _p(f).I _q(g) whenever f; g are forms of degree d over F in x _l, …, x _p; …, x _q respectively, and where (f ⊥ g)(x _l, …, x _p+q) = f(x ₁, …, x _p,) + g (x _p+1, …, x _p+q). It is well-known that the family of discriminants of the quadratic forms is additive. We prove that in odd degree d each invariant in an additive family must be a constant. We also give an example in each even degree d of a nontrivial family of invariants of the forms of degree d. The proofs depend on the symbolic method for representing invariants of a form, which we review.     

一类pnm阶群的构造

本文研究了一类特殊的pnm阶有限群的构造.利用求解数论同余方程的方法和群的扩张理论,得到了具有m阶循环正规子群,其补子群为循环群的Pnm阶有限群的构造及相关的计数定理.     

小康判别函数的建立与分析

近几年，国内在研究小康水平的定量指标分析中，应用较多的有综合评分法、层次分析法和模糊识别法。这几种方法比单项指标更全面、灵活，但在权数的确定上都难以克服人为因素的影响。当指标之间存在两个或两个以上的高度相关时，对问题的研究总存在一定的局限性。为了科学地研究小康问题，本文基于判别分析，建立城市小康的判别系数，来综合评价出我国城市小康水平的定量标准。     

有限Abel群上子集等价类的计数

设G为一有限Abel群，|G|=v,D1,D2是G的两个子集，如果存在t∈Zv,(t,v)=1,s∈G使D1＝tD2 s，则称D1与D2是等价的。文中给出了G的k-子集等价类的计数公式，同时也给出了G的的所有子集等价类的计数公式。     

ESTIMATES OF N-FUNCTION AND m-FUNCTION OF MEROMORPHIC SOLUTIONS OF SYSTEMS OF COMPLEX DIFFERENCE EQUATIONS

We apply Nevanlinna theory of the value distribution of meromorphic functions to study the properties of Nevanlinna counting function and proximity function of meromorphic solutions of a type of systems of complex difference equations. Our results can give estimates on the proximity function and the counting function of solutions of systems of difference equations. This implies that solutions have a relatively large number of poles. It extend some result concerning difference equations to the systems of difference equations.