On some continuity and differentiability properties of paths of Gaussian processes

The following path properties of real separable Gaussian processes ξ with parameter set an arbitrary interval are established. At every fixed point the paths of ξ are continuous, or differentiable, with probability zero or one. If ξ is measurable, then with probability one its paths have essentially the same points of continuity and differentiability. If ξ is measurable and not mean square continuous or differentiable at every point, then with probability one its paths are almost nowhere continuous or differentiable, respectively. If ξ harmonizable or if it is mean square continuous with stationary increments, then its paths are absolutely continuous with probability one if and only if ξ is mean square differentiable; also mean square differentiability of ξ implies path differentiability with probability one at every fixed point. If ξ is mean square differentiable and stationary, then on every interval with probability one its paths are either differentiable everywhere or nondifferentiable on countable dense subsets. Also a class of harmonizable processes is determined for which of the following are true: (i) with probability one paths are either continuous or unbounded on every interval, and (ii) mean square differentiability implies that with probability one on every interval paths are either differentiable everywhere or nondifferentiable on countable dense subsets.     

Merging asymptotic expansions for semistable random variables

Merging asymptotic expansions are established for distribution functions from the domain of geometric partial attraction of a semistable law. The length of the expansion depends on the exponent of the semistable law and on the characteristic function of the underlying distribution. We obtain sufficient conditions for the quantile function in order to get real infinite asymptotic expansion. The results are generalizations of the existing theory in the stable case.     

Series representation for semistable laws and their domains of semistable attraction

If the centered and normalized partial sums of an i.i.d. sequence of random variables converge in distribution to a nondegenerate limit then we say that this sequence belongs to the domain of attraction of the necessarily stable limit. If we consider only the partial sums which terminate atk _n wherek _n+1 ck _n then the sequence belongs to the domain of semistable attraction of the necessarily semistable limit. In this paper, we consider the case where the limiting distribution is nonnormal. We obtain a series representation for the partial sums which converges almost surely. This representation is based on the order statistics, and utilizes the Poisson process. Almost sure convergence is a useful technical device, as we illustrate with a number of applications.This research was supported by a research scholarship from the Deutsche Forschungsgemeinschaft (DFG).     

On the sample path properties of mixed Poisson processes

The mixed Poisson process has been widely used in financial engineering for modeling arrival of events that cluster in time, as it has strictly stationary and positively correlated increments. However, we show that, surprisingly, the sample autocovariance and autocorrelation of the increments of a mixed Poisson process converge to zero almost surely as the sample size goes to infinity. Consequently, the sample autocovariance or autocorrelation cannot be used in the method of moments for parameter estimation of mixed Poisson processes.     

On characterisation of Markov processes via martingale problems

It is well-known that well-posedness of a martingale problem in the class of continuous (or r.c.l.l.) solutions enables one to construct the associated transition probability functions. We extend this result to the case when the martingale problem is well-posed in the class of solutions which are continuous in probability. This extension is used to improve on a criterion for a probability measure to be invariant for the semigroup associated with the Markov process. We also give examples of martingale problems that are well-posed in the class of solutions which are continuous in probability but for which no r.c.l.l. solution exists.     

Picard groups of the moduli spaces of semistable sheaves I

We compute the Picard group of the moduli spaceU′ of semistable vector bundles of rankn and degreed on an irreducible nodal curveY and show thatU′ is locally factorial. We determine the canonical line bundles ofU′ andU _L ′, the subvariety consisting of vector bundles with a fixed determinant. For rank 2, we compute the Picard group of other strata in the compactification ofU′.     

Large deviations ordering of point processes in some queueing networks

Given a stochastic ordering between point processes, say that a p.p. N is smooth if it is less than the Poisson process with the same average intensity for this ordering. In this article we investigate whether initially smooth processes retain their smoothness as they cross a network of FIFO ·/D/1 queues along fixed routes. For the so-called strong variability ordering we show that point processes remain smooth as they proceed through a tandem of quasi-saturated (i.e., loaded to 1) M+·/D/1 queues. We then introduce the Large Deviations ordering, which involves comparison of the rate functions associated with Large Deviations Principles satisfied by the point processes. For this ordering, we show that smoothness is retained when the processes cross a feed-forward network of unsaturated ·/D/1 queues. We also examine the LD characteristics of a deterministic p.p. at the output of an M+·/D/1 queue. This revised version was published online in June 2006 with corrections to the Cover Date.     

Iteratively sampling scheme for stochastic optimization with variable number sample path

Optimal search methods are proposed for solving optimization problems with analytically unobtainable objectives. This paper proposes a method by incorporating sampling schemes into the directional direct search with variable number sample path and investigates its effectiveness in solving stochastic optimization problems. We also explore the conditions on sample sizes at each iteration under which the convergence in probability can be guaranteed. Finally, a set of benchmark problems are numerically tested to show the effectiveness in different sampling schemes.     

On a functional equation generalizing the class of semistable distributions

With ?(p),p≥0 the Laplace-Stieltjes transform of some infinitely divisible probability distribution, we consider the solutions to the functional equation ?(p-e ^?pβΠ _i=1 ^m ?^γi (c _i p) for somem≥1,c _i>0, γ _i >0,i=1., …,m, β ε ®. We supply its complete solutions in terms of semistable distributions (the ones obtained whenm=1). We then show how to obtain these solutions as limit laws (r → ∞) of normalized Poisson sums of iid samples when the Poisson intensity λ(r) grows geometrically withr.     

Stochastic processes with sample paths in reproducing kernel Hilbert spaces

A theorem of M. F. Driscoll says that, under certain restrictions, the probability that a given Gaussian process has its sample paths almost surely in a given reproducing kernel Hilbert space (RKHS) is either or . Driscoll also found a necessary and sufficient condition for that probability to be .

Doing away with Driscoll's restrictions, R. Fortet generalized his condition and named it nuclear dominance. He stated a theorem claiming nuclear dominance to be necessary and sufficient for the existence of a process (not necessarily Gaussian) having its sample paths in a given RKHS. This theorem - specifically the necessity of the condition - turns out to be incorrect, as we will show via counterexamples. On the other hand, a weaker sufficient condition is available.

Using Fortet's tools along with some new ones, we correct Fortet's theorem and then find the generalization of Driscoll's result. The key idea is that of a random element in a RKHS whose values are sample paths of a stochastic process. As in Fortet's work, we make almost no assumptions about the reproducing kernels we use, and we demonstrate the extent to which one may dispense with the Gaussian assumption.

     

How to optimize discrete-event systems from a single sample path by the score function method

We present a method for deriving theoptimal solution of a class of mathematical programming problems, associated with discrete-event systems and in particular with queueing models, while using asingle sample path (single simulation experiment) from the underlying process. Our method, called thescore function method, is based on probability measure transformation derived from the efficient score process and generating statistical counterparts to the conventional deterministic optimization procedures (e.g. Lagrange multipliers, penalty functions, etc.). Applications of our method to optimization of various discrete-event systems are presented, and numerical results are given.Research supported by the L. Edelstein Research Fund at the Technion-Israel Institute of Technology.     

lp-值Gauss过程样本轨道性质的注记

crements The a.s. sample path properties for ι^p-valued Gaussian processes with stationary increments under some more general conditions are established.     

Sample path properties of stochastic integrals,and stochastic differentiation

Let B be a Brownian motion, and X = H.B be a stochastic integral of B. We give conditions on the smoothness of the process H which imply that if Ms a singular point of the sample path of B (ω) (such as a local maximum, a slow point, or a fast point) then t is also a singular point of X (ω). In the final section we give an application to stochastic differential equations     

On the law of the iterated logarithm for Gaussian processes

We present some optimal conditions for the compact law of the iterated logarithm of a sequence of jointly Gaussian processes in different situations. We also discuss the local law of the iterated logarithm for Gaussian processes indexed by arbitrary index sets, in particular for self-similar Gaussian processes. We apply these results to obtain the law of the iterated logarithm for compositions of Gaussian processes. Research partially supported by NSF Grant DMS-93-02583.     

On the spectrum of Markov semigroups via sample path large deviations

The essential spectral radius of a sub-Markovian process is defined as the infimum of the spectral radiuses of all local perturbations of the process. When the family of rescaled processes satisfies sample path large deviation principle, the spectral radius and the essential spectral radius are expressed in terms of the rate function. The paper is motivated by applications to reflected diffusions and jump Markov processes describing stochastic networks for which the sample path large deviation principle has been established and the rate function has been identified while essential spectral radius has not been calculated.     

可加Lévy 过程的轨道渐近性质

本文主要研究有限个相互独立的从属过程之和的样本轨道的渐近性质. 给出了样本轨道在零点 附近和无穷远处的渐近增长率的上下极限, 并且得出了在零点附近渐近增长率的一致下极限.     

A class of risk processes with reserve-dependent premium rate: sample path large deviations and importance sampling

Let (X(t)) be a risk process with reserve-dependent premium rate, delayed claims and initial capital u. Consider a class of risk processes {(X ^ε (t)): ε > 0} derived from (X(t)) via scaling in a slow Markov walk sense, and let Ψ_ε(u) be the corresponding ruin probability. In this paper we prove sample path large deviations for (X ε (t)) as ε → 0. As a consequence, we give exact asymptotics for log Ψ_ε(u) and we determine a most likely path leading to ruin. Finally, using importance sampling, we find an asymptotically efficient law for the simulation of Ψ_ε(u). AMS Subject Classifications 60F10, 91B30 This work has been partially supported by Murst Project “Metodi Stocastici in Finanza Matematica”     

Convergence of probabilities for the second order monadic properties of a random mapping

We prove that the probability of each second order monadic property of a random mapping u_n converges as n→∞. © 1997 John Wiley & Sons, Inc. Random Struct. Alg., 11 , 277–295, 1997     

Green function,capacity and sample path properties for a class of hypoelliptic diffusion processes

Summary This paper presents a detailed probabilistic study of a class of hypoelliptic diffusion processes in ₃ including the so-called Brownian motion on the Heisenberg group. We obtain precise estimates for the Green function of the process and the capcity of small compact sets. These estimates are applied to various sample path properties such as the existence of double points or the asymptotic behavior of the volume of a small tubular neighborhood of the path. Differences with the elliptic case are emphasized.Dedicated to Klaus Krickeberg on the occasion of his 60th birthday     

On the properties of small sample of GM(1,1) model

In grey prediction modeling, the more samples selected the more errors. This paper puts forward new explanations of “incomplete information and small sample” of grey systems and expands the suitable range of grey system theory. Based on the geometric sequence, it probes into the influence on the relative errors by selecting the different sample sizes. The research results indicate that to the non-negative increasing monotonous exponential sequence, the more samples selected, the more average relative errors. To the non-negative decreasing monotonous exponential sequence, a proper sample number exists that has the least average relative error. When the initial value of the sequence of raw data of new information GM(1,1) model changes, the development coefficient remains unchanged. The segmental correction new information GM(1,1) model (SNGM) can obviously improve the simulation accuracy. It puts forward the mathematic proofs that the small sample usually has more accuracy than the large sample when establishing GM(1,1) model in theory.