Probabilistic characterizations of the Generalized Domain of Attraction of the multivariate normal law

For independent identically distributed random vectors,X _i , we give necessary and sufficient probabilistic conditions for their common distribution to belong to the Generalized Domain of Attraction of the multivariate normal law. The first condition says that after projecting onto any direction, , the sum of squares, _i ¹⁼¹ X _i , ², properly normalized, converges to one in probability uniformly over the unit sphere. The second condition says that max X _i , ²/ _n ⁱ⁼¹ X _i , ² converges to zero in probability uniformly over the unit sphere.     

Minimaxity and admissibility of the product limit estimator

In this article we examine the minimaxity and admissibility of the product limit (PL) estimator under the loss function% MathType!MTEF!2!1!+-% feaafeart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqaqpepeea0xe9qqVa0l% b9peea0lb9sq-JfrVkFHe9peea0dXdarVe0Fb9pgea0xa9pue9Fve9% Ffc8meGabaqaciGacaGaaeqabaWaaeaaeaaakeaacaWGmbGaaiikai% aadAeacaGGSaGabmOrayaajaGaaiykaiabg2da9maapeaabaGaaiik% aiaadAeacaGGOaGaamiDaiaacMcaaSqabeqaniabgUIiYdGccqGHsi% slceWGgbGbaKaacaGGOaGaamiDaiaacMcacaGGPaWaaWbaaSqabeaa% caaIYaaaaOGaamOramaaCaaaleqabaaccmGae8xSdegaaOGaaiikai% aadshacaGGPaGaaiikaiaaigdacqGHsislcaWGgbGaaiikaiaadsha% caGGPaGaaiykamaaCaaaleqabaGaeqOSdigaaOGaamizaiaadEfaca% GGOaGaamiDaiaacMcaaaa!5992!\[L(F,\hat F) = \int {(F(t)} - \hat F(t))^2 F^\alpha (t)(1 - F(t))^\beta dW(t)\].To avoid some pathological and uninteresting cases, we restrict the parameter space to ={F: F(ymin) }, where (0, 1) and y ₁,...y,_n are the censoring times. Under this set up, we obtain several interesting results. When y ₁=···=y _n, we prove the following results: the PL estimator is admissible under the above loss function for , {–1, 0}; if n=1, ==–1, the PL estimator is minimax iff dW ({y})=0; and if n2, , {–1, 0}, the PL estimator is not minimax for certain ranges of . For the general case of a random right censorship model it is shown that the PL estimator is neither admissible nor minimax. Some additional results are also indicated.Partially supported by the Governor's Challenge Grant.Part of the work was done while the author was visiting William Paterson College.     

Uniformly summing sets of operators on spaces of vector–valued continuous functions

Let X, Y be Banach spaces. We say that a set is uniformly p–summing if the series is uniformly convergent for whenever (x_n) belongs to . We consider uniformly summing sets of operators defined on a -space and prove, in case X does not contain a copy of c₀, that is uniformly summing iff is, where T (φ x) = (T^#φ) x for all and x∈X. We also characterize the sets with the property that is uniformly summing viewed in . Received: 1 July 2005     

On the convergence of the LJ search method

The convergence of the Luus-Jaakola search method for unconstrained optimization problems is established.Notation E _n Euclideann-space - f Gradient off(x) - ² f Hessian matrix - (·) ^T Transpose of (·) - I Index set {1, 2, ...,n} - [x _i1 ^*(j) ] Point around which search is made in the (j + 1)th iteration, i.e., [x _1l ^*(j) ,x _2l ^*(j) ,...,x _n1 ^*(j) ] - r _i ⁽ⁱ⁾ Range ofx _il ^*(i) in the (j + 1)th iteration - l ₁ min_i {r _i ⁽⁰⁾ } - l ₂ min_i {r _i ⁽⁰⁾ } - A _j Region of search in thejth iteration, i.e., {x E _n:x _il ^*(j-1) –0.5r _i ^(j-1) x _ix _il ^*(j-1) +0.5r _i ^(j-1) ,i I} - S _j Closed sphere with center origin and radius _j - Reduction factor in each iteration - 1– - (·) Gamma function Many discussions with Dr. S. N. Iyer, Professor of Electrical Engineering, College of Engineering, Trivandrum, India, are gratefully acknowledged. The author has great pleasure to thank Dr. K. Surendran, Professor, Department of Electrical Engineering, P.S.G. College of Technology, Coimbatore, India, for suggesting this work.     

A Remark on Bernstein,Prokhorov, Bennett,Hoeffding, and Talagrand Inequalities

Let X ₁,..., X_n be independent random variables such that {X_j 1}=1 and E X _j=0 for all j. We prove an upper bound for the tail probabilities of the sum M _n=X₁+...+ X_n. Namely, we prove the inequality {M _nx} 3.7 {S_n x}, where S _n=₁+...+ _n is a sum of centered independent identically distributed Bernoulli random variables such that E S _n ² =ME M _n ² and {_k=1}=E S _n ² /(n+E S _n ² ) for all k (we call a random variable Bernoulli if it assumes at most two values). The inequality holds for x at which the survival function x{S _nx} has a jump down. For remaining x, the inequality still holds provided that we interpolate the function between the adjacent jump points linearly or log-linearly. If necessary, in order to estimate {S _nx} one can use special bounds for binomial probabilities. Up to the factor at most 2.375, the inequality is final. The inequality improves the classical Bernstein, Prokhorov, Bennett, Hoeffding, Talagrand, and other bounds.     

On the Rate of Clustering to the Strassen Set for Increments of the Uniform Empirical Process

We obtain outer rates of clustering in the functional laws of the iterated logarithm of Deheuvels and Mason⁽¹¹⁾ and Deheuvels,⁽⁷⁾ which describe local oscillations of empirical processes. Considering increment sizes a _n 0 such that na _n and na _n(log n)^–7/3 we show that the sets of properly rescaled increment functions cluster with probability one to the _n-enlarged Strassen ball in B(0, 1) endowed with the uniform topology, where _n 0 may be chosen so small as (log (1/a _n) + log log n)^–2/3 for any sufficiently large . This speed of coverage is reduced for smaller a _n.     

Certain embedding theorems

We obtain necessary and sufficient conditions such that, for f(x) from LP(0, 1), the integral ₀ ¹ ¦f (x)¦qdx (0<p<1,p<q<p(1 –p)^–1) is convergent, or for f LP[0, 1] for all p 1, the integral ₀ ¹ e¦f(x)¦dx is convergent.Translated from Matematicheskie Zametki, Vol. 19, No. 2, pp. 187–200, February, 1976.     

Hitting time of a half-line by two-dimensional random walk

We consider the probability that a two-dimensional random walk starting from the origin never returns to the half-line {(x₁,x₂)|x₁0,x₂=0} before time n. It is proved that for aperiodic random walk with mean zero and finite 2+(>2)-th absolute moment, this probability times n^1/4 converges to some positive constant c^* as . We show that c^* is expressed by using the characteristic function of the increment of the random walk. For the simple random walk, this expression gives Mathematics Subject Classification (2000):60G50, 60E10     

Random Logistic Maps. I

Let {C _i} ₀ be a sequence of independent and identically distributed random variables with vales in [0, 4]. Let {X _n} ₀ be a sequence of random variables with values in [0, 1] defined recursively by X _n+1=C _n+1 X _n(1–X _n). It is shown here that: (i) E ln C ₁<0X _n0 w.p.1. (ii) E ln C ₁=0X _n0 in probability (iii) E ln C ₁>0, E |ln(4–C ₁)| such that (0, 1)=1 and is invariant for {X _n}. (iv) If there exits an invariant probability measure such that {0}=0, then E ln C ₁>0 and – ln(1–x) (dx)=E ln C ₁. (v) E ln C ₁>0, E |ln(4–C ₁)|< and {X _n} is Harris irreducible implies that the probability distribution of X _n converges in the Cesaro sense to a unique probability distribution on (0, 1) for all X ₀0.     

Three theorems on ρ* random fields

This paper generalizes results by Bradley.⁽³⁾ Suppose that for 1=1,2,...X _k ¹ :k ^d is a centered, weakly stationary ^*-mixing random field, and suppose lim_l Cov(X ₀ ¹ ,x _k ¹ ) exists, anyk ^d . Then the successive spectral densities converge uniformly to a continuous function. For a sequence of strictly stationary random fields that are uniformly ^*-mixing and satisfy a indeberg condition, a CLT is proved for sequences of sums from the fields. This result is then applied: given a centered strictly stationary ^*-mixing random field whose probability density and joint densities are continuous, then a kernel estimator for the probability density obeys the CLT.     

Approximation Schemes for Scheduling Jobs with Common Due Date on Parallel Machines to Minimize Total Tardiness

The problem of scheduling n nonpreemptive jobs having a common due date d on m, m 2, parallel identical machines to minimize total tardiness is studied. Approximability issues are discussed and two families of algorithms {A } and {B } are presented such that (T ⁰ – T*)/(T* + d) holds for any problem instance and any given > 0, where T* is the optimal solution value and T ⁰ is the value of the solution delivered by A or B . Algorithms A and B run in O(n ^2m / ^m–1) and O(n ^m+1/ ^m ) time, respectively, if m is a constant. For m = 2, algorithm A can be improved to run in O(n ³/) time.     

A Note on Primes and Goldbach Numbers in Short Intervals

Let J(N, H) be the Selberg integral and E(x, T) the error term in Kaczorowski-Perelli's weighted form of the classical explicit formula. We prove that the estimate J(N, H) = o(H2 N) is connected with an appropriate estimate of _N ^2N| E(x,T)² dx, uniformly for H and T in some ranges. Moreover, assuming a suitable bound for _N ^2N| E(x,T)|² dx, we also obtain, for all sufficiently large N and H (log N)11/12, that every interval [N,N + H] contains H Goldbach numbers.     

A self normalized law of the iterated logarithm for random walk in random scenery

LetX,X _i ,i1, be a sequence of i.i.d. random vectors in ^d . LetS _o=0 and, forn1, letS _n =X ₁+...+X _n . LetY,Y(), ^d , be i.i.d. -valued random variables which are independent of theX _i . LetZ _n =Y(S _o )+...+Y(S _n ). We will callZ _n arandom walk in random scenery.In this work, we consider the law of the iterated logarithm for random walk in random sceneries. Under fairly general conditions, we obtain arandomly normalized law of the iterated logarithm.Supported in part by NSF Grants DMS-85-21586 and DMS-90-24961.     

Efficiency and the uniform linear minorization of convex functions

LetX be a Banach space, and let {f _i:iI} be a family of proper lower semicontinuous convex functions defined onX, each of whose epigraphs meets a fixed bound subset ofX×. We say that {f _i:iI} is uniformly linearly minorized if there exists a positive scalar such that for alliI andxX, we havef _i(x)–(1+x). We present two very different characterizations of uniform linear minorization for such a family. Using one of these, we show that either strong or weak epi-convergence of a sequence of convex functions at some point in the effective domain of the limit implies, uniform linear minorization for the entire sequence.With 1 Figure     

Uniform dimension results for Gaussian random fields

Let X = {X(t), t ∈ ℝ ^N } be a Gaussian random field with values in ℝ ^d defined by

Limit of Solutions of a SDE with a Large Drift Driven by a Poisson Random Measure

We consider a sequence of {X _n} of R ^d-valued processes satisfying a stochastic differential equation driven by a Brownian motion and a compensated Poisson random measure, with _n ~ _n with a large drift. Let be a m-dimensional submanifold (m<d), where F vanishes. Then under some suitable growth conditions for _n ~ _n, and some conditions for F, we show that dist(X _n, )0 before it exits any given compact set, that is, the large drift term forces X _n close to . And if the coefficients converge to some continuous functions, any limit process must actually stay on and satisfy a certain stochastic differential equation driven by Brownian motion and white noise.     

Estimates of the tail of the stationary density function of certain nonlinear autoregressive processes

We consider the Markov chainX _n+1=T(X _n )+ _n , where { _n ;n1} is a ^d -valued random sequence of independent identically distributed random variables, and the functionT: ^{d d} is measurable and satisfies a suitable growth condition. Under certain conditions involvingT and the probability distribution of _n , we show that this Markov chain is ergodic. Moreover, we obtain sharp upper bounds for the tail of the corresponding stationary probability density function. In our proofs, we make use of the Leray-Schauder fixed-point theorem.     

Exponential Convergence in Probability for Empirical Means of Brownian Motion and of Random Walks

Given a Brownian motion (B _t) _t0 in R ^d and a measurable real function f on R ^d belonging to the Kato class, we show that 1/t ₀ ^t f(B _s ) ds converges to a constant z with an exponential rate in probability if and only if f has a uniform mean z. A similar result is also established in the case of random walks.     

Small Ball Probabilities Around Random Centers of Gaussian Measures and Applications to Quantization

Let be a centered Gaussian measure on a separable Hilbert space (E, ). We are concerned with the logarithmic small ball probabilities around a -distributed center X. It turns out that the asymptotic behavior of –log (B(X,)) is a.s. equivalent to that of a deterministic function _R (). These new insights will be used to derive the precise asymptotics of a random quantization problem which was introduced in a former article by Dereich, Fehringer, Matoussi, and Scheutzow.⁽⁸⁾     

An improved asymptotic analysis of the expected number of pivot steps required by the simplex algorithm

Leta ₁ ...,a _m be i.i.d. points uniformly on the unit sphere in ⁿ ,m n 3, and letX:= {x ⁿ |a _i ^T x1} be the random polyhedron generated bya ₁, ...,a _m . Furthermore, for linearly independent vectorsu, in ⁿ , letS _u , (X) be the number of shadow vertices ofX inspan(u,). The paper provides an asymptotic expansion of the expectation value¯S _n,m := _in4 ¹ E(S _u, ) for fixedn andm .¯S _n,m equals the expected number of pivot steps that the shadow vertex algorithm — a parametric variant of the simplex algorithm — requires in order to solve linear programming problems of type max u ^T ,xX, if the algorithm will be started with anX-vertex solving the problem max ^T ,x X. Our analysis is closely related to Borgwardt's probabilistic analysis of the simplex algorithm. We obtain a refined asymptotic analysis of the expected number of pivot steps required by the shadow vertex algorithm for uniformly on the sphere distributed data.