This is a rambling review of what, with a few notable and significant exceptions, has been a rather dormant area for over a decade. It concentrates on the septuagenarian problem of finding good approximations for the excursion probability $P\{sup_{t \in T}X_t \geq \lambda\}$, where $\lambda$ is large, $X$ is a Gaussian, or “Gaussian-like,” process over a region $T \subset \Re^N$, and, generally, $N > 1$. A quarter of a century ago, there was a flurry of papers out of various schools linking this problem to the geometrical properties of random field sample paths. My own papers made the link via Euler characteristics of the excursion sets $\{t \in T: X_t \geq \lambda\}$. A decade ago, Aldous popularized the Poisson clumping heuristic for computing excursion probabilities in a wide variety of scenarios, including the Gaussian. Over the past few years, Keith Worsley has been the driving force behind the computation of many new Euler characteristic functionals, primarily driven by applications in medical imaging. There has also been a parallel development of techniques in the astrophysical literature. Meanwhile, somewhat closer to home, Hotelling’s 1939 “tube formulas” have seen a renaissance as sophisticated statistical hypothesis testing problems led to their reapplication toward computing excursion probabilities, and Sun and others have shown how to apply them in a purely Gaussian setting. The aim of the present paper is to look again at many of these results and tie them together in new ways to obtain a few new results and, hopefully, considerable new insight. The “Punchline of this paper,”which relies heavily on a recent result of Piterbarg, is given in Section 6.6: “In computing excursion probabilities for smooth enough Gaussian random fields over reasonable enough regions, the expected Euler characteristic of the corresponding excursion sets gives an approximation, for large levels, that is accurate to as many terms as there are in its expansion.”

A recent paper by Harrison and Van Mieghem explained in general mathematical terms how one forms an “equivalent workload formulation” of a Brownian network model. Denoting by $Z(t)$ the state vector of the original Brownian network, one has a lower dimensional state descriptor $W(t) = MZ(t)$ in the equivalent workload formulation, where $M$ can be chosen as any basis matrix for a particular linear space. This paper considers Brownian models for a very general class of open processing networks, and in that context develops a more extensive interpretation of the equivalent workload formulation, thus extending earlier work by Laws on alternate routing problems. A linear program called the static planning problem is introduced to articulate the notion of “heavy traffic ” for a general open network, and the dual of that linear program is used to define a canonical choice of the basis matrix $M$. To be specific, rows of the canonical $M$ are alternative basic optimal solutions of the dual linear program. If the network data satisfy a natural monotonicity condition, the canonical matrix $M$ is shown to be nonnegative, and another natural condition is identified which insures that $M$ admits a factorization related to the notion of resource pooling.

We consider a symmetric network composed of $N$ links, each with capacity $C$. Calls arrive according to a Poisson process, and each call concerns $L$ distinct links chosen uniformly at random. If each of these links has free capacity, the call is held for an exponential time; otherwise it is lost. The semiexplicit stationary distribution for this process is similar to a Gibbs measure: it involves a normalizing factor, the partition function, which is very difficult to evaluate.We let $N$ go to infinity and keep fixed the rate of call attempts concerning any link. We use asymptotic combinatorics and recent techniques involving the law of large numbers to obtain the asymptotic equivalent for the logarithm of the partition function and then the large deviation principle for the empirical measure of the occupancies of the links. We give an explicit formula for the rate function and examine its properties.

Using duality, an expansion is found for the transition function of the reversible $K$-allele diffusion model in population genetics. In the neutral case, the expansion is explicit but already known. When selection is present, it depends on the distribution at time $t$ of a specified $K$-type birth-and-death process starting at “infinity.” The latter process is constructed by means of a coupling argument and characterized as the Ray process corresponding to the Ray–Knight compactification of the $K$-dimensional nonnegative-integer lattice.

We consider a reaction-diffusion process with two components, on the grid $\mathbb{Z}$. This process had been introduced by Durrett and Levin to describe a two-species interaction. We prove the process admits hydrodynamic limits, first with a technique based on correlation functions, then with the method of relative entropy plus coupling.

We consider the parabolic Anderson problem $\partial_{t}u = \kappa\Delta u + \xi(x)u$ on $\mathbb{R}_+ \times \mathbb{R}^d$ with initial condition $u(0, x) = 1$. Here $\xi(\cdot)$ is a random shift-invariant potential having high $\delta$-like peaks on small islands. We express the second-order asymptotics of the $p$th moment $(p \in [1, \infty))$ of $u(t,0)$ as $t \to \infty$ in terms of a variational formula involving an asymptotic description of the rescaled shapes of these peaks via their cumulant generating function. This includes Gaussian potentials and high Poisson clouds.

Let $W \geq 0$ be a random variable with $EW = 1$, and let $Z^{(r)} (r \geq 2)$ be the limit of a Mandelbrot’s martingale, defined as sums of product of independent random weights having the same distribution as $W$, indexed by nodes of a homogeneous $r$-ary tree. We study asymptotic properties of $Z^{(r)}$ as $r \to \infty$: we obtain a law of large numbers, a central limit theorem, a result for convergence of moment generating functions and a theorem of large deviations. Some results are extended to the case where the number of branches is a random variable whose distribution depends on a parameter $r$.

Starting at time 0, unit-length intervals arrive and are placed on the positive real line by a unit-intensity Poisson process in two dimensions; the left endpoints of intervals appear at the rate of 1 per unit time per unit distance. An arrival is accepted if and only if, for some given x , the interval is contained in $[o, x]$ and overlaps no interval already accepted. This stochastic, on-line interval packing problem generalizes the classical parking problem, the latter corresponding only to the absorbing states of the interval packing process, where successive packed intervals are separated by gaps less than or equal to 1 in length. In earlier work,the authors studied the distribution of the number of intervals accepted during 0 t .This paper is concerned with the vacant intervals (gaps)between consecutive packed intervals.Let p t y be the limit x 8of the fraction of the gaps at time t which are at most y in length.We prove that $$p(t,y)= \begin{cases} \frac{2\int^t_0(1−e^{−vy})\beta(v)dv}{\alpha(t)}, \qquad\qquad\qquad y ≤ 1,\\ p(t,1)+\frac{(1−\exp(−t(y−1)))t\beta(t)}{\alpha(t)}, \quad y > 1,\end{cases} $$ where $\alpha(t) = \int_0^t \beta(v)dv, \beta(t) = \exp[-2 \int_0^t((1-e^{-v})/v)dv]$. We briefly discuss the recent importance acquired by interval packing models in connection with resource-reservation systems. In these applications, our vacant intervals correspond to times between consecutive reser- vation intervals. The results of this paper improve our understanding of the fragmentation of time that occurs in reservation systems.

Let random points $X_1,\dots, X_n$ be sampled in strict sequence from a continuous product distribution on Euclidean $d$-space. At the time $X_j$ is observed it must be accepted or rejected. The subsequence of accepted points must increase in each coordinate. We show that the maximum expected length of a subsequence selected is asymptotic to $\gamma n^{1/(d+1)}$ and give the exact value of $\gamma$. This extends the $\sqrt{2n}$ result by Samuels and Steele for $d = 1$.

We study an exchange economy comprising $n$ agents,where the excess demands by the agents are random variables. We show that under certain conditions the set of price equilibria is nonempty. We also prove a theorem concerning the convergence of the random price equilibria toward the price equilibria of an associated “expectation economy.”

We study the stochastic control problem of maximizing expected utility from terminal wealth, when the wealth process is subject to shocks produced by a general marked point process; the problem of the agent is to derive the optimal allocation of his wealth between investments in a nonrisky asset and in a (costly) insurance strategy which allows “lowering” the level of the shocks. The agent’s optimization problem is related to a suitable dual stochastic control problem in which the constraint on the insurance strategy disappears. We establish a general existence result for the dual problem as well as the duality between both problems. We conclude by some applications in the context of power (and logarithmic) utility functions and linear insurance premium which show, in particular, the existence of two critical values for the insurance premium: below the lower critical value, agents prefer to be completely insured, whereas above the upper critical value they take no insurance.

We prove that the Hölder singularities of random lacunary wavelet series are chirps located on random fractal sets. We determine the Hausdorff dimensions of these singularities, and the a.e. modulus of continuity of the series. Lacunary wavelet series thus turn out to be a new example of multifractal functions.