\(L^p(\mu)\)-estimation of tangent maps (Q1304640)

In numerical experiments with dynamical systems, one often needs to estimate the derivative (the tangent map) of a diffeomorphism. This is necessary, for example, in the computations of Lyapunov exponents. When a direct method of calculating the derivative is not available, the Eckmann-Ruelle algorithm can be used. It is based on fitting (in the \(L^p\) norm) a linear function (the tangent map) to experimentally computed images of several points in a small ball. Here the authors analyze conditions under which the best \(L^p\) fit gives reliable estimates of the tangent map.