Search

The embedding theorem of Takens, and its extensions, have provided the theoretical underpinning for a wide range of investigations of time series derived from nonlinear dynamical systems. The theorem applies when the dynamical system is sampled uniformly in time. There has, however, been increasing interest in situations where observations on the system are not uniform in time, and in particular where the consist of a series of inter-event (`interspike') intervals. Sauer has provided an embedding theorem for the case where these intervals are generated by an integrate-and-fire mechanism. Here we prove several embedding theorems pertaining to non-uniform sampling. We consider two situations: in the first, observations consist of the values of some function on the state space of the system, with the times between the successive observations being given by another function—the sampling interval function; in the second, sampling times are generated in the same way, but now the observations consist only of the intersample intervals. We prove embedding theorems both when the sampling interval function is allowed to be rather general, and when it corresponds to integrate-and-fire sampling. We point out that non-uniform sampling might lead to better reconstructions than uniform sampling, for certain kinds of time series.

We describe methods for the study of topological properties of the invariant manifolds of experimental dynamical systems. We explain how to compute such invariants as the Euler characteristic and Betti numbers using time series data, and suggest a number of potential applications.

This paper introduces a new class of models of digital communications channels. Physically, these models take account of the digital nature of the input. Mathematically, they are iterated function systems. As a consequence of making explicit assumptions about the role of discreteness in the models, it is possible to make general statements about the behaviour of these channels without needing to assume that they are linear. We provide the mathematical background necessary to understand the behaviour of these models and prove a number of results about their observability. We also provide a number of examples intended to demonstrate their connection with linear state-space models, and to suggest how the nonlinear theory might be developed towards applications.

It has been asserted in the literature that the low pass filtering of time series data may lead to erroneous results when calculating attractor dimensions. Here we prove that finite order, non-recursive filters do not have this effect. In fact, a generic, finite order, non-recursive filter leaves invariant all the quantities that can be estimated by using embedding techniques such as the method of delays.

Recent work on modelling digital channels using iterated function systems suggests a general approach to the theory of signal processing in digital communications which uses so-called delay methods developed for deterministic nonlinear timeseries analysis. Here we make the connection between this work and the more conventional approach to digital communications by casting linear channel models as iterated function systems and showing how the use of delay methods gives a nice connection with the theory of observability in the control of linear systems.

In biological systems, chemical reactions often take place in complex spatial environments. For example, the translation of m-RNA to produce protein within eukaryotic cells takes place within the extremely crowded cytoplasmic environment and appears to require the spatial coordination of many translation factors. It is important, therefore, to understand the transport processes within such an environment. While there is growing interest in both experimental and computational studies of such environments, it is also important to develop suitable mathematical models. Here, as an example of such a model, we study a reaction-diffusion equation defined on the Sierpinski gasket. Both experimental and computational studies of analogous systems have shown power law behaviour and associated deviations from mass action kinetics. The analysis presented here allows us to distinguish the roles of the fractal domain and of the discreteness of molecular interactions in producing this effect. Indeed, we show that the fractal domain alone is insufficient.

An approach to finding codes for use in direct sequence spread spectrum communications systems is described. It is based upon an analogy between codes having auto- and cross-correlation properties desirable for spread spectrum systems, and certain dynamical systems encountered in ergodic theory called systems with Lebesgue spectrum. Such systems are associated with collections of orthogonal functions and these functions can be used to generate collections of time series with zero cross-correlation functions. To generate codewords we must use truncated versions of these time series, for which the cross-correlations are no longer precisely zero: these truncated sequences correspond to periodic orbits of the dynamical system. The method for finding a code from a suitable periodic orbit is described, and an example, using a simple dynamical system known as the doubling map, is worked through in some detail.

The embedding theorem forms a bridge between the theory of nonlinear dynamical systems and the analysis of experimental time series. This memorandum describes the theorem and gives a detailed account of its proof. The necessary differential topology is briefly reviewed, and then a proof of the theorem is presented; this proof follows broadly the argument of Takens, although it differs in some details. Some extensions to the theorem, which facilitate its use in applications, are described. The memo concludes with a brief discussion of what the theorem implies about time series, viewed as the raw material for signal processing algorithms.

We present a new embedding theorem for time series, in the spirit of Takens's theorem, but requiring multivariate signals. Our result is part of a growing body of work that extends the domain of geometric time series analysis to some genuinely stochastic systems---including such natural examples as x_{j+1} = \phi(x_j) + \eta_j where \phi is some fixed map and the \eta_j are i.i.d. random displacements.

Takens’ Embedding Theorem forms the basis of virtually all approaches to the analysis of time series generated by nonlinear deterministic dynamical systems. It typically allows us to reconstruct an unknown dynamical system which gave rise to a given observed scalar time series simply by constructing a new state space out of successive values of the time series. This provides the theoretical foundation for many popular techniques, including those for the measurement of fractal dimensions and Liapunov exponents, for the prediction of future behaviour, for noise reduction and signal separation, and most recently for control and targeting. Current versions of Takens’ Theorem assume that the underlying system is autonomous (and noise-free). Unfortunately this is not the case for many real systems. In a previous paper, one of us showed how to extend Takens’ Theorem to deterministically forced systems. Here, we use similar techniques to prove a number of delay embedding theorems for arbitrarily and stochastically forced systems. As a special case, we obtain embedding results for Iterated Functions Systems, and we also briefly consider noisy observations.