Fractional Brownian Density Process and Its Self-Intersection Local Time of Order k

The fractional Brownian density process is a continuous centered Gaussian ( ^d )-valued process which arises as a high-density fluctuation limit of a Poisson system of independent d-dimensional fractional Brownian motions with Hurst parameter H. ( ( ^d ) is the space of tempered distributions). The main result proved in the paper is that if the intensity measure of the (initial) Poisson random measure on ^d is either the Lebesgue measure or a finite measure, then the density process has self-intersection local time of order k 2 if and only if Hd < k/(k – 1). The latter is also the necessary and sufficient condition for existence of multiple points of order k for d-dimensional fractional Brownian motion, as proved by Talagrand¹². This result extends to a non-Markovian case the relationship known for (Markovian) symmetric -stable Lévy processes and their corresponding density processes. New methods are used in order to overcome the lack of Markov property. Other properties of the fractional Brownian density process are also given, in particular the non-semimartingale property in the case H 1/2, which is obtained by a general criterion for the non-semimartingale property of real Gaussian processes that we also prove.     

A Note on Transient Gaussian Fluid Models

In this paper we study the distribution of the supremum over interval [0,T] of a centered Gaussian process with stationary increments with a general negative drift function. This problem is related to the distribution of the buffer content in a transient Gaussian fluid queue Q(T) at time T, provided that at time 0 the buffer is empty. The general theory is illustrated by detailed considerations of different cases for the integrated Gaussian process and the fractional Brownian motion. We give asymptotic results for P(Q(T)>x) and P(sup _0tT Q(t)>x) as x.     

On the martingale problem for Banach space valued stochastic differential equations

LetE denote a real separable Banach space and letZ=(Z(t, f) be a family of centered, homogeneous, Gaussian independent increment processes with values inE, indexed by timet0 and the continuous functionsf:[0,t] E. If the dependence ont andf fulfills some additional properties,Z is called a gaussian random field. For continuous, adaptedE-valued processesX a stochastic integral processY = ₀ ^. Z(t, X)(dt) is defined, which is a continuous local martingale with tensor quadratic variation[Y] = ₀ ^. Q(t, X)dt, whereQ(t, f) denotes the covariance operator ofZ(t, f).Y is called a solution of the homogeneous Gaussian martingale problem, ifY = ₀ ^. Z(t, Y)(dt). Such solutions occur naturally in connection with stochastic differential equations of the type (D):dX(t)=G(t, X) dt+Z(t, X)(dt), whereG is anE-valued vector field. It is shown that a solution of (D) can be obtained by a kind of variation of parameter method, first solving a deterministic integral equation only involvingG and then solving an associated homogeneous martingale problem.     

On Some Reliability Applications of Rice's Formula for the Intensity of Level Crossings

Let X be a stationary process with absolutely continuous sample paths. If is finite and if the distribution of X(0) is absolutely continuous, then, for almost all u, the crossing intensity () of the level u by X(t) is given by the generalized Rice's formula . The classical Rice's formula for (), which is valid for a fixed level u, , holds under more restrictive technical conditions that can be difficult to check in applications. In this paper it is shown that often in practice the weaker form of Rice's formula (valid for almost all u) is sufficient. Three engineering problems are discussed; prediction of fatigue life time; computing the average stress at slams and analysis of crest height of sea waves.     

Géométrie Orthogonale Non Symétrique et Congruences Quadratiques (Nonsymmetric Orthogonal Geometry and Quadratic Congruences)

This paper is devoted to the study of quadratic congruences, which appear in the theory of exceptional bundles on projective spaces. A quadratic congruence associates to each point P of a projective space a quadratic in containing P. We study several geometric constructions of quadratic congruences and try to classify them.     

Global ellipsoidal approximations and homotopy methods for solving convex analytic programs

This paper deals with some problems of algorithmic complexity arising when solving convex programming problems by following the path of analytic centers (i.e., the trajectory formed by the minimizers of the logarithmic barrier function). We prove that in the case ofm convex quadratic constraints we can obtain in a simple constructive way a two-sided ellipsoidal approximation for the feasible set (intersection ofm ellipsoids), whose tightness depends only onm. This can be used for the early identification of those constraints which are active at the optimum, and it also explains the efficiency of Newton's method used as a corrector when following the central path. Various parametrizations of the central path are studied. This also leads to an extrapolation (predictor) algorithm which can be regarded as a generalization of the method of conjugate gradients.This research was supported by the Deutsche Forschungsgemeinschaft as part of a major research project Anwendungsbezogene Optimierung und Steuerung.     

Adaptive boundary control of stochastic linear distributed parameter systems described by analytic semigroups

A stochastic adaptive control problem is formulated and solved for some unknown linear, stochastic distributed parameter systems that are described by analytic semigroups. The control occurs on the boundary. The highest-order operator is assumed to be known but the lower-order operators contain unknown parameters. Furthermore, the linear operators of the state and the control on the boundary contain unknown parameters. The noise in the system is a cylindrical white Gaussian noise. The performance measure is an ergodic, quadratic cost functional. For the identification of the unknown parameters a diminishing excitation is used that has no effect on the ergodic cost functional but ensures sufficient excitation for strong consistency. The adaptive control is the certainty equivalence control for the ergodic, quadratic cost functional with switchings to the zero control.This research was partially supported by NSF Grants ECS-9102714, ECS-9113029, and DMS-9305936.     

Modified proximal point algorithm for extended linear-quadratic programming

Extended linear-quadratic programming arises as a flexible modeling scheme in dynamic and stochastic optimization, which allows for penalty terms and facilitates the use of duality. Computationally it raises new challenges as well as new possibilities in large-scale applications. Recent efforts have been focused on the fully quadratic case ([15] and [23]), while relying on the fundamental proximal point algorithm (PPA) as a shell of outer iterations when the problem is not fully quadratic. In this paper, we focus on the nonfully quadratic cases by proposing some new variants of the fundamental PPA. We first construct a continuously differentiable saddle function S(u, v) through infimal convolution in such a way that the optimal primal-dual pairs of the original problem are just the saddle points of S(u, v) on the whole space. Then the original extended linear-quadratic-programming problem reduces to solving the nonlinear equation S(u, v)=0. We then embed the fundamental PPA and some of its previous variants in the framework of a Newton-like iteration for this equation. After revealing the local quadratic structure of S near the solution, we derive new extensions of the fundamental PPA. In numerical tests, the modified iteration scheme based on the quasi-Newton update formula outperforms the fundamental PPA considerably.     

Parametrization via secant length and application to path following

Summary A possible way for parametrizing the solution path of the nonlinear systemH(u)=0, H: ^{n+1 n} consists of using the secant length as parameter. This idea leads to a quadratic constraint by which the parameter is introduced. A Newton-like method for computing the solution for a given parameter is proposed where the nonlinear system is linearized at each iterate, but the quadratic parametrizing equation is exactly satisfied. The localQ-quadratic convergence of the method is proved and some hints for implementing the algorithm are givenDedicated to Professor Lothar Collatz on the occasion of his 75th birthday     

Gaussian sample functions and the Hausdorff dimension of level crossings

Summary Let X _t be a real Gaussian process with stationary increments, mean 0, _t ² =E[(X _s+t–X _s)²] If _t ² behaves like t as t 0, 0<<1, the graph of a.e. sample function will have Hausdorff dimension 2 -. This leads one to feel that the set of zeros of X _t should have Hausdorff dimension 1 -. This is shown to be true provided the process is stationary and satisfies additional assumptions.     

Joint distributions of numbers of occurrences of a discrete pattern and weak convergence of an empirical process for the pattern

For a {0, 1}-pattern of finite length, an empirical process is introduced in order to describe the number of overlapping occurrences of the pattern at each level t[0,1] in a sequence of the corresponding indicators of i.i.d. [0, 1]-valued observations of length n. A method for obtaining the exact finite-dimensional distributions of the empirical process is given. The weak convergence of the process to a Gaussian process in D[0,1] as n tends to infinity is also established. The limiting process depends on the given pattern. The exact covariance function is compared with the asymptotic covariance function in a numerical example.     

Photon statistics in the multi-quantum interaction of a generalized two-level system with a strongly squeezed magnetic field

Interaction of squeezed light and a two-level system with a degenerate excited level is considered. The dependence of the relative fluctuations of magnetic field density, , as a function of the squeezing parameters and the process photon number, n, was found. The value of is shown to be weakly dependent on the method of squeezing for low-intensity fields and small n; for increasing n and field intensity, the value of is much larger under phase than amplitude squeezing.Translated from Teoreticheskaya i Matematicheskaya Fizika, Vol. 105, No. 3, pp. 471–477, December, 1995.     

Evolution in a Gaussian Random Field

We consider an evolution process in a Gaussian random field V(q) with the mean ‹V(q)› = 0 and the correlation function W(|q – q|) ‹V(q)V(q)›, where q ^d and d is the dimension of the Euclidean space ^d . For the value ‹G(q,t;q ₀)›, t > 0, of the Green's function of the evolution equation averaged over all realizations of the random field, we use the Feynman–Kac formula to establish an integral equation that is invariant with respect to a continuous renormalization group. This invariance property allows using the renormalization group method to find an asymptotic expression for ‹G(q,t;q ₀)› as |q – q ₀| and t .     

On the convergence of the method of analytic centers when applied to convex quadratic programs

This work examines the method of analytic centers of Sonnevend when applied to solve generalized convex quadratic programs — where also the constraints are given by convex quadratic functions. We establish the existence of a two-sided ellipsoidal approximation for the set of feasible points around its center and show, that a simple (zero order) algorithm starting from an initial center of the feasible set generates a sequence of strictly feasible points whose objective function values converge to the optimal value. Concerning the speed of convergence it is shown that an upper bound for the gap in between the objective function value and the optimal value is reduced by a factor of with iterations wherem is the number of inequality constraints. Here, each iteration involves the computation of one Newton step. The bound of Newton iterations to guarantee an error reduction by a factor of in the objective function is as good as the one currently given forlinear programs. However, the algorithm considered here is of theoretical interest only, full efficiency of the method can only be obtained when accelerating it by some (higher order) extrapolation scheme, see e.g. the work of Jarre, Sonnevend and Stoer.This work was supported by the Deutsche Forschungsgemeinschaft, Schwerpunktprogramm für anwendungsbezogene Optimierung und Steuerung.     

Extremes of Gaussian Processes with Maximal Variance near the Boundary Points

Let X(t), t[0,1], be a Gaussian process with continuous paths with mean zero and nonconstant variance. The largest values of the Gaussian process occur in the neighborhood of the points of maximum variance. If there is a unique fixed point t₀ in the interval [0,1], the behavior of P{sup_t[0,1] X(t)>u} is known for u. We investigate the case where the unique point t₀ = t_u depends on u and tends to the boundary. This is reasonable for a family of Gaussian processes X_u(t) depending on u, which have for each u such a unique point t_u tending to the boundary as u. We derive the asymptotic behavior of P{sup_t[0,1] X(t)>u}, depending on the rate as t_u tends to 0 or 1. Some applications are mentioned and the computation of a particular case is used to compare simulated probabilities with the asymptotic formula. We consider the exceedances of such a nonconstant boundary by a Ornstein-Uhlenbeck process. It shows the difficulties to simulate such rare events, when u is large.     

Tight Gaussian 4-Designs

A Gaussian t-design is defined as a finite set X in the Euclidean space ℝⁿ satisfying the condition: for any polynomial f(x) in n variables of degree at most t, here α is a constant real number and ω is a positive weight function on X. It is easy to see that if X is a Gaussian 2e-design in ℝⁿ, then . We call X a tight Gaussian 2e-design in ℝⁿ if holds. In this paper we study tight Gaussian 2e-designs in ℝⁿ. In particular, we classify tight Gaussian 4-designs in ℝⁿ with constant weight or with weight . Moreover we classify tight Gaussian 4-designs in ℝⁿ on 2 concentric spheres (with arbitrary weight functions).     

Simultaneous Distributions of Space-Time Wave Characteristics in a Gaussian Sea

A simultaneous pair (L, T) is defined, representing crest length and crest period of the same random wave, i.e. the length and duration of the half of the wave that contains the crest. The simultaneous density in the ergodic sense of (L, T) is derived and evaluated for a homogeneous Gaussian spatio-temporal random sea model. Other names of this type of distribution are Palm, intensity, and long-run distribution. The density is compared to the deterministic relation given by the dispersion relation for deep water: L=T ² g/. A two-dimensional sea model parametrized by time and a single space coordinate is used, based on a directional spectrum. Only the case of deep water is considered. Furthermore, the crest height associated with the (L, T)-pair is studied. The ergodic distribution of (L, T) for significant waves is compared to that for all waves. All results are verified through comparison to simulated observations, but no real data are considered.     

Quadratic kernel-free non-linear support vector machine

A new quadratic kernel-free non-linear support vector machine (which is called QSVM) is introduced. The SVM optimization problem can be stated as follows: Maximize the geometrical margin subject to all the training data with a functional margin greater than a constant. The functional margin is equal to W ^T X + b which is the equation of the hyper-plane used for linear separation. The geometrical margin is equal to . And the constant in this case is equal to one. To separate the data non-linearly, a dual optimization form and the Kernel trick must be used. In this paper, a quadratic decision function that is capable of separating non-linearly the data is used. The geometrical margin is proved to be equal to the inverse of the norm of the gradient of the decision function. The functional margin is the equation of the quadratic function. QSVM is proved to be put in a quadratic optimization setting. This setting does not require the use of a dual form or the use of the Kernel trick. Comparisons between the QSVM and the SVM using the Gaussian and the polynomial kernels on databases from the UCI repository are shown.     

Central limit theorems for stochastic processes under random entropy conditions

Summary Necessary and sufficient conditions are found for the weak convergence of the row sums of an infinitesimal row-independent triangular array ( _nj ) of stochastic processes, indexed by a set S, to a sample-continuous Gaussian process, when the array satisfies a random entropy condition, analogous to one used by Giné and Zinn (1984) for empirical processes. This entropy condition is satisfied when S is a class of sets or functions with the Vapnik-ervonenkis property and each _nj (f)fdnj is of the form njc for some reasonable random finite signed measure v _nj. As a result we obtain necessary and sufficient conditions for the weak convergence of (possibly non-i.i.d.) partial-sum processes, and new sufficient conditions for empirical processes, indexed by Vapnik-ervonenkis classes. Special cases include Prokhorov's (1956) central limit theorem for empirical processes, and Shorack's (1979) theorems on weighted empirical processes.Research supported by an NSF Postdoctoral Fellowship, grant no. MCS83-111686     

Amenability,Poincaré series and quasiconformal maps

Summary Any coveringYX of a hyperbolic Riemann surfaceX of finite area determines an inclusion of Teichmüller spaces Teich(X)Teich(Y). We show this map is an isometry for the Teichmüller metric if the covering isamenable, and contracting otherwise. In particular, we establish <1 for classical Poincaré series (Kra's Theta conjecture).The appendix develops the theory of geometric limits of quadratic differentials, used in this paper and a sequel.Research partially supported by an NSF Postdoctoral Fellowship