Approximating Markov Chains: What and why
- 990 Moose Hill Road, Guilford, Connecticut 06437 (United States)
Much of the current study of dynamical systems is focused on geometry (e.g., chaos and bifurcations) and ergodic theory. Yet dynamical systems were originally motivated by an attempt to {open_quote}{open_quote}solve,{close_quote}{close_quote} or at least understand, a discrete-time analogue of differential equations. As such, numerical, analytical solution techniques for dynamical systems would seem desirable. We discuss an approach that provides such techniques, the approximation of dynamical systems by suitable finite state Markov Chains. Steady state distributions for these Markov Chains, a straightforward calculation, will converge to the true dynamical system steady state distribution, with appropriate limit theorems indicated. Thus (i) approximation by a computable, linear map holds the promise of vastly faster steady state solutions for nonlinear, multidimensional differential equations; (ii) the solution procedure is unaffected by the presence or absence of a probability density function for the {ital attractor}, entirely skirting singularity, fractal/multifractal, and renormalization considerations. The theoretical machinery underpinning this development also implies that under very general conditions, steady state measures are weakly continuous with control parameter evolution. This means that even though a system may change periodicity, or become chaotic in its limiting behavior, such statistical parameters as the mean, standard deviation, and tail probabilities change continuously, not abruptly with system evolution. {copyright} {ital 1996 American Institute of Physics.}
- OSTI ID:
- 401071
- Report Number(s):
- CONF-950730-; ISSN 0094-243X; TRN: 96:029448
- Journal Information:
- AIP Conference Proceedings, Vol. 375, Issue 1; Conference: 3. technical conference on nonlinear dynamics (chaos) and full spectrum processing, Mystic, CT (United States), 10-13 Jul 1995; Other Information: PBD: Jun 1996
- Country of Publication:
- United States
- Language:
- English
Similar Records
Avalanche dynamics in evolution, growth, and depinning models
Heat conduction in one-dimensional chains and nonequilibrium Lyapunov spectrum