Motivated partly by the resurgence of neural computation research, and partly by advances in device technology, there has been a recent increase of interest in analog, continuous-time computation. However, while special-case algorithms and devices are being developed, relatively little work exists on the general theory of continuous-time models of computation. In this paper, we survey the existing models and results in this area, and point to some of the open research questions. 1 Introduction After a long period of oblivion, interest in analog computation is again on the rise. The immediate cause for this new wave of activity is surely the success of the neural networks "revolution", which has provided hardware designers with several new numerically based, computationally interesting models that are structurally sufficiently simple to be implemented directly in silicon. (For designs and actual implementations of neural models in VLSI, see e.g. [30, 45]). However, the more fundamental...

The result of J. Mather on the existence of trajectories with unbounded energy for time periodic Hamiltonian systems on a torus is generalized to a class of multidimensional Hamiltonian systems with Hamiltonian polynomial in momenta. It is assumed that the leading homogeneous term of the Hamiltonian is autonomous and the corresponding Hamiltonian system has a hyperbolic invariant torus possessing a transversal homoclinic trajectory. Under certain Melnikov type condition, the existence of trajectories with unbounded energy is proved. Instead of the variational methods of Mather, a geometrical approach based on KAM theory and the Poincaré-Melnikov method is used. This makes it possible to study a more general class of Hamiltonian systems, but requires additional smoothness assumptions on the Hamiltonian.

Dedicated to the memory of

this paper, we present the weakly damped Schrodinger equation, which is a system generated by a dispersive equation with weak damping.

Given an analytic integrable Hamiltonian with three or more degrees of freedom, we construct, arbitrarily close to it, an analytic perturbation with transition chains whose lengths only depend on the unperturbed Hamiltonian. Then we deduce that the perturbed system has Arnold di usion. We provide the technical details of the tools we use.

this paper is to discuss the denition of hyperbolic invariant tori of Hamiltonian systems. Hyperbolic tori were rst studied by Poincare. Arnold [1] named hyperbolic tori as whiskered and used them in his famous example of the phenomenon that is now called \Arnold diusion". KAM-type results on the conservation of hyperbolic tori under small perturbations were obtained by Gra [16] and Zehnder [27]. In the recent years the notion of a hyperbolic torus was widely used in literature [25, 22, 7], but, as far as we know, there is no universally accepted denition. Usually a denition of hyperbolic torus is given by presenting some normal form of the Hamiltonian in its neighborhood. The goal of this note is to give an invariant (coordinate independent) denition of a hyperbolic turus and derive some consequences. This paper is mainly of the methodological nature and doesn't contain really new results. However, we hope that it will be useful for clarifying the subject.

The main goal of this dissertation is to show the existence of global cross sections for certain classes of Anosov flows. Let \Phi be a C codimension one Anosov flow on a compact Riemannian manifold M of dimension greater than three. Verjovsky conjectured that \Phi admits a global cross section and we affirm this conjecture in the following three cases: 1) if the sum, E , of the strong stable and strong unstable bundle of \Phi is Lipschitz; 2) if E is `-Holder continuous for all ` ! 1 and \Phi preserves volume; 3) if the center stable distribution of \Phi is of class C for all ` ! 1 and \Phi preserves volume. We note that 1) and 3) generalize the results of Ghys from [Gh3]. For showing 1), we needed to prove a natural generalization of the theorem of Frobenius on integrability of distributions which are only Lipschitz. We also

A time-periodic Hamiltonian system on a cotangent bundle of a compact manifold with Hamiltonian strictly convex and superlinear in the momentum is studied. A hyperbolic Diophantine nondegenerate invariant torus N is said to be minimal if it is a Peierls set in the sense of the Aubry-Mather theory. We prove that N has an infinite number of homoclinic orbits. For any family of homoclinic orbits the first and last intersection point with the boundary of a tubular neighborhood U of N define sets in U . If no family of minimal homoclinics defines a non-contractible set in U , we obtain an infinite number of multibump homoclinic, periodic and chaotic orbits. The proof is based on a combination of variational methods of Mather and a generalization of Shilnikov's lemma to invariant tori.

Semi-hyperbolic dynamical systems generated by Lipschitz mappings are shown to be exponentially expansive, locally at least, and explicit rates of expansion are determined. The result is applicable to nonsmooth noninvertible systems such as those with hysteresis e#ects as well as to classical systems involving hyperbolic di#eomorphisms. AMS Subject Classification 58F15 1 Introduction Complicated dynamical behaviour is often a consequence of the expansivity of a dynamical system, a concept that has been formulated quite simply: no two distinct trajectories can remain forever within a prescribed threshhold of each other. Expansive systems have, not surprisingly, been investigated intensively for many years (cf. [7], [8], [9]), particularly in regard to their entropy and structural stability. Frequently considered examples of expansive systems involve hyperbolic di#eomorphisms, such as Smale's horseshoe mapping, in which the rate of separation of trajectories is in fact exponential, at...

: Digital image analysis appears to be more and more relevant to the study of physical phenomena involving fluid motion, and of their evolution over time. In that context, 2D deformable motion analysis is one of the most important issues to be investigated. The interpretation of such deformable 2D flow fields can generally be stated as the characterization of linear models provided that first order approximations are considered in an adequate neighborhood of so-called singular points, where the velocity becomes null. This paper describes an efficient method, based on a statistical approach, which explicitly addresses these problems, and allows us to locate, characterize and track such singular points in an image sequence. It does not require the prior computation of the velocity field. The method has been validated by experiments carried out with synthetic and real examples corresponding to a meteorological image sequence. In fact, the described approach can be of interest in different...