Small ball probabilities for Gaussian processes with stationary increments under Hölder norms (Q1890742): Difference between revisions

The article is devoted to the estimation of the probability of small balls for the Gaussian processes in Hölder norms. For the centered Gaussian process with the stationary increments \(X\) on \([0;1]\) the following estimates are proved \[ \text{exp}\{-c(\beta)/x\} \leq P\{\lambda_ f(X) \leq \sigma(x)/f(x)\} \leq \text{exp}\{-[ x/2 ] \cdot \ln (1/ \Phi (2/\sqrt {4 - \theta}))\}, \] where \(\sigma^ 2(t) = EX^ 2(t)\), \(t > 0\), \(X(0) = 0\), \(f\) is the function which generates Hölder norm \(\lambda_ f\), \(\Phi\) is the standard normal distribution, \(\sigma^ 2(2x) \leq \theta \sigma^ 2(x)\), \(x \in [0; 1/2 ]\), \(\theta \in [0; 4 ]\). Such estimation allows to obtain the speed of the convergence in the LIL for the fractional Brownian motion for the different points of the cluster set.