Large Fluctuations of Randomly Perturbed Dynamical Systems
Abstract
A finite-dimensional dynamical system that is subject to random
perturbations, such as a white-noise driving force, may
fluctuate far from its attractor or attractors. In fact,
the random perturbations may induce fluctuations between
domains of attraction. Large fluctuations of this sort may
be analysed either probabilistically (via the theory of
large deviations) or analytically (via WKB-type approximations
and matched asymptotic approximations). This talk will explain
how very precise asymptotic results may be obtained by the latter
technique. In the weak-noise limit of a noise-perturbed
system, it turns out that the most probable fluctuational
dynamics are described by a variational principle. This
permits accurate computation of the frequency with which such
fluctuations occur, in appropriate asymptotic regimes.
Applications include fluctuation phenomena in statistical physics,
chemical physics, and in telecommunications (the overflow of
data buffers).