Vol 214, No 4 (2016)

Two approaches that allow to construct a probabilistic representation of a generalized solution of the Cauchy problem for a system of quasilinear parabolic equations are proposed. The system under consideration describes a population dynamics model for a prey-predator population. The stochastic problem associated with this parabolic system is presented in two forms, which give a way to derive the required probabilistic representation. Bibliography: 16 titles.

The aim of the paper is to study a service system model introduced by I. Kaj and M. Taqqu. A limit theorem for the process of integral workload on the service system is proved. This theorem generalizes the corresponding result of I. Kaj and M. Taqqu, because the weak convergence in the Skorokhod space is established.

The Large Deviation Principle is proved for the conditional probabilities of moderate deviations of weighted empirical bootstrap measures with respect to a fixed empirical measure. Using this LDP for the problem of calculation of moderate deviation probabilities of differentiable statistical functionals, it is shown that the importance sampling based on influence function is asymptotically efficient.

A generalization of a service system model introduced by I. Kaj and M. Taqqu is considered. Unlike the original model, the unnatural assumption on independence between the duration and required resources quantity of a service process is dropped. A number of limit theorems for the process of integral workload is presented. Among the considered limit processes are the Winer process, fractional Brownian motion, and stable Lévy process.

Properties of generalized infinitely divisible distributions with Lévy measure \( \varLambda (dx)=\frac{g(x)}{x^{1+\upalpha}}dx, \) α ∈ (2, 4) ∪ (4, 6) are studied. Such a measure is a signed one and, hence, is not a probability measure. It is proved that in some sense these signed measures are the limit measures for the distributions of the sums of independent random variables. Bibliography: 6 titles

The paper considers a connection between weighted norm inequalities for the Hilbert transform with matrix valued weights and an estimating problem. A connection of the vector Muckenhoupt condition on the spectral density of the stationary noise and a possibility to transform a difficult estimating problem to another well-studied problem is established. Bibliography: 12 titles

A one-dimensional diffusion process is considered. The characteristic operator of this process is assumed to be a linear differential operator of the second order with negative coefficient at the term with zero derivative. Such an operator determines the measure of a Markov diffusion process with break (the first interpretation), and also the measure of a semi-Markov diffusion process with final stop (the second interpretation). Under the second interpretation, the existence of the limit of the process at infinity (the final point) is characterized. This limit exists on any interval almost surely with respect to the conditional measure generated by the condition that the process never leaves this interval. The distribution of the final point expressed in terms of two fundamental solutions of the corresponding ordinary differential equation, and also the distribution of the instant final stop are derived. A homogeneous process is considered as an example.