Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · nlin/0101029

Predictability: a way to characterize Complexity

G. Boffetta·M. Cencini·M. Falcioni·A. Vulpiani
View original source ↗Read source document ↗Export metadata ↓

Abstract

Different aspects of the predictability problem in dynamical systems are reviewed. The deep relation among Lyapunov exponents, Kolmogorov-Sinai entropy, Shannon entropy and algorithmic complexity is discussed. In particular, we emphasize how a characterization of the unpredictability of a system gives a measure of its complexity. Adopting this point of view, we review some developments in the characterization of the predictability of systems showing different kind of complexity: from low-dimensional systems to high-dimensional ones with spatio-temporal chaos and to fully developed turbulence. A special attention is devoted to finite-time and finite-resolution effects on predictability, which can be accounted with suitable generalization of the standard indicators. The problems involved in systems with intrinsic randomness is discussed, with emphasis on the important problems of distinguishing chaos from noise and of modeling the system. The characterization of irregular behavior in systems with discrete phase space is also considered.

Explore related subjects

nlin.CDcond-mat

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

G. Boffetta, M. Cencini, M. Falcioni, A. Vulpiani. 2001-01-17. Predictability: a way to characterize Complexity. https://doi.org/10.1016/s0370-1573(01)00025-4

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

When chaos meets thermodynamics: from Fermi acceleration to thermalisation

A particle colliding with a moving wall provides one of the simplest mechanical settings in which a small system can exchange energy with its surroundings. In a chaotic time-dependent billiard, repeated elastic impacts may produce Fermi acceleration, with the characteristic kinetic energy growing without bound. This result poses an instructive thermodynamic puzzle: a physical wall maintained at finite temperature should act as a reservoir and drive a dilute gas toward a stationary temperature rather than heat it indefinitely. We use this puzzle to connect nonlinear dynamics with elementary thermodynamics. Under suitable decorrelation assumptions, collision sequences generate effective diffusion in velocity space and the square-root growth associated with normal diffusion. We then introduce two ingredients absent from the conservative accelerator: phase randomisation, which simplifies the statistics of energy exchange, and dissipation, which transfers energy back to unresolved degrees of freedom of the boundary. Their competition replaces unbounded acceleration by a stationary kinetic-energy scale. A recurrence for the second velocity moment yields a stationary kinetic-temperature scale $T_\infty=(mA^2/8k_B)(1+γ)/(1-γ)$, where $A$ measures the boundary forcing and $γ$ is a restitution coefficient. We compare relaxation in collision number with a simple Fourier-inspired heat-transfer model in physical time, emphasising that the two clocks are not interchangeable during transients. The resulting picture shows how fluctuations, dissipation, coarse graining and relaxation can produce an effective thermal state from chaotic dynamics while also clarifying the limits of the thermodynamic analogy.

nlin.CD↗
arXiv2026

Physics-Inspired Optimizers beyond the Edge of Stability: Period-Two Orbits and Delayed Instability

Gradient descent with momentum can be viewed as a particle moving with friction in the loss landscape. In deep learning the learning rate is usually large, so training is a discrete dynamical system: in many full-batch experiments the sharpness grows to the optimizer's stability limit, the edge of stability (EoS), and training continues there. Physics-inspired optimizers change this motion; Velocity-Regularized Adam (VRAdam), for example, shrinks the step at high speed. We study kinetic maps, in which the step is the gradient of a kinetic energy of the momentum, and ask what they do beyond the EoS. (i) In a reduced model, a kinetic map creates a symmetric period-two orbit beyond the edge, strictly linearly stable if and only if the map's elasticity (the log-log slope of step length versus momentum) lies between 0 and 1; on it, training sits at an effective edge of stability. VRAdam's damper fails above twice the critical learning rate, where its kinetic energy is not convex; the Legendre dual of VRAdam's own Lagrangian is convex and passes at every learning rate. (ii) Because the sharpness rises gradually, the loss of stability is delayed, as in Neishtadt's theory of slowly swept bifurcations. In a network with a controlled sweep, the dynamics match this theory without fitted parameters, including the memory of where the sweep started and its erasure by noise. (iii) With momentum, reaching the oscillation also requires elasticity at large momentum: maps that bound the step have competing period-four and chaotic attractors already at beta1 = 0.9, while the cubic-Legendre map reached the oscillation from every sampled start up to beta1 = 0.99. Experiments on an exact quadratic problem, a full-batch MLP and ResNet-32 support the theory; on ResNet-18, much of the apparent benefit of the maps is reproduced by matching the effective learning rate.

nlin.CD↗
arXiv2026

Three-Wave Forced Hasegawa--Wakatani Model: Reducing Chaos with Zonal Flows

We investigate the nonlinear dynamics of resistive drift-wave turbulence with a three-wave truncation of the Hasegawa--Wakatani equations, in which the background density gradient either is fixed or evolves self-consistently, fed by an external particle source $P$ (such as neutral beam injection) and eroded by the turbulent particle flux. Two triads that differ only by a small tilt of the long-wavelength mode are compared, in one this mode is zonal and cannot be driven by the gradient, in the other it is a weakly driven drift wave. Period maps in the plane of the adiabaticity parameter $C$ and the drive show that the zonal triad is almost always regular, whereas the non-zonal triad becomes chaotic. The feedback of the flux on the gradient amplifies this contrast: the fraction of aperiodic states grows considerably in the non-zonal triad and remains negligible in the zonal one. Along a line of constant $C$ at low adiabaticity, the non-zonal triad with a fixed gradient enters chaos through torus breakdown with intermittency and leaves it through an interior crisis and a period-doubling cascade. With the self-consistent gradient, a Neimark--Sacker bifurcation leads to a phase-locked torus that breaks down abruptly, with hysteresis, and chaos then persists up to the largest drive, with the gradient driven intermittently above the threshold of the long-wavelength mode. None of these routes to chaos exists when that mode is zonal.

nlin.CD↗