Estimates of the supremum distribution for a certain class of random processes (Q1335980): Difference between revisions

There exist many papers devoted to a study of the distribution of the supremum of random processes. The most significant results have been obtained for the Gaussian random processes, for certain classes of Markov processes, etc. Recently, the entropy methods, efficient in the Gaussian case, have been applied to finding the exponential estimates for certain classes of random processes which resemble, in a certain sense, the Gaussian processes. The idea of this paper is close to that of \textit{V. A. Dmitrovskij} [Theory Probab. Math. Stat. 25, 169-180 (1982); translation from Teor. Veroyatn. Mat. Stat. 25, 154-164 (1981; Zbl 0465.60043)], where the exponential estimates were obtained for the distribution of the supremum of the general pre-Gaussian processes. Below, we indicate a subclass of the class of pre-Gaussian processes, for which the exponential estimates obtained earlier can be improved. It is worth noting that this subclass contains processes that can be represented as quadratic forms of the Gaussian processes or as stochastic integrals of the processes with independent increments. Note that in Section 2 we study the entropy characteristics connected not with the semimetrics induced by random processes but with more general functionals.