Correlation dimension for iterated function systems

Given positive integers \(N\) and \(k\), parameters \(a_{ij}\) and \(p_{ij}\) with \(0\leq p_{ij}\leq 1- a_{ij}\) \((i= 1,\dots, k,\;j=1,\dots, N)\) let \(Q= [0, 1)^N\) and define maps \(T_i: Q\to Q\) \((i= 1,\dots, k)\) by \(T_i(x_1,\dots, x_N)= (a_{ij} x_j+ p_{ij}\); \(j= 1,\dots, N)\). Assume that all \(a_{ij}\) are less than \(1/2\) and consider the skinny baker's map given by \[ Q\times [0,1)\ni (x_1,\dots, x_N, x_{n+ 1})\mapsto (T_i(x_1,\dots, x_N), kx_{N+ 1}\pmod 1), \] where \(i\) is the least integer greater than \(kx_{N+ 1}\). Putting \(T(S)= \bigcup^k_{i= 1}T_i(S)\) for all sets \(S\subset Q\) denote by \(A\) the attractor of \(T\), i.e., \(A= \bigcap^\infty_{n= 1} T^n(Q)\). Finally, let \(D_2(A)\) be the correlation dimension of \(A\), that is \[ D_2(A)= \sup\Biggl\{s\in (0,+ \infty): \int_{A\times A} {1\over| x-y|^2} d\mu\times \mu(x, y)<\infty\Biggr\}, \] where \(| x-y|\) is the distance between \(x\) and \(y\) in \(\mathbb{R}^N\) and \(\mu(S)\) is the natural measure of \(A\) generated by \(T\) (the averaged fraction of the points of the \(T\)-trajectory in \(A\)). The main result states that for fixed \(k\) and the parameters \(a_{ij}\) the dimension \(D_2(A)\) typically do not depend on the numbers \(p_{ij}\), i.e., has the same value for almost all \(p_{ij}\) with respect to Lebesgue measure matrices. As a corollary, the authors have answered affirmatively a question of Kaplan and Yorke on the relation between information and Lyapunov dimensions.

Recommendations

Cites work

Cited in

(15)

This page was built for publication: Correlation dimension for iterated function systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4336540)