Hierarchical training of neural networks and prediction of chaotic time series

Physics Letters APublished 1 August 1991

J. Deppisch, H.-U. Bauer, T. Geisel

DOI10.1016/0375-9601(91)90340-e

Source URL

Citations23

SJR quartileQ2

SJR score0.46

SNIP0.81

TL;DR

A new procedure for hierarchical training of multilayer perceptrons to outputs of high precision achieves a dramatic increase in accuracy, e.g. by three orders of magnitude, and can reduce training time considerably.

Abstract

We present a new procedure for hierarchical training of multilayer perceptrons to outputs of high precision. It achieves a dramatic increase in accuracy, e.g. by three orders of magnitude, and can reduce training time considerably. The method is applied to the prediction of chaotic systems where we obtain the optimum error evolution for iterated predictions as well as a substantial reduction of the absolute prediction error.

Keywords

Computer SciencePhysics and Astronomy

Physical Review LettersGeometry from a Time Series

3,889 Citations1980Norman H. Packard, James P. Crutchfield +2 more

Physics Letters AAn equation for continuous chaos

3,885 Citations1976Otto E. Rössler

Physical Review LettersPredicting chaotic time series

1,945 Citations1987J. Doyne Farmer, John J. Sidorowich

An error estimate is presented for this forecasting technique for chaotic data, and its effectiveness is demonstrated by applying it to several examples, including data from the Mackey-Glass delay differential equation, Rayleigh-Benard convection, and Taylor-Couette flow.

NatureNonlinear forecasting as a way of distinguishing chaos from measurement error in time series

1,943 Citations1990George Sugihara, Robert M. May

An approach is presented for making short-term predictions about the trajectories of chaotic dynamical systems, applied to data on measles, chickenpox, and marine phytoplankton populations, to show how apparent noise associated with deterministic chaos can be distinguished from sampling error and other sources of externally induced environmental noise.

Europhysics Letters (EPL)Optimal Embeddings of Chaotic Attractors from Topological Considerations

204 Citations1991Wolfgang Liebert, Klaus Pawelzik +1 more

A new method is introduced to obtain optimal delay coordinates for data from chaotic dynamic systems by determining simultaneously the minimal necessary embedding dimension as well as the proper delay time to achieve optimal reconstructions of attractors.

Physica D Nonlinear PhenomenaOn determining the dimension of chaotic flows

159 Citations1981Harold Froehling, James P. Crutchfield +3 more

NatureCalculating the rate of loss of information from chaotic time series by forecasting

149 Citations1991David J. Wales

A method for estimating from such forecasting the largest Liapunov exponent of the dynamics, which provides a measure of how chaotic the system is—that is, how rapidly information is lost from the system.

Physical Review AUnstable periodic orbits and prediction

47 Citations1991Klaus Pawelzik, Heinz Georg Schuster

Physical Review ANonlinear dynamics of feedback multilayer perceptrons

11 Citations1990H.-U. Bauer, T. Geisel

The nonlinear dynamics of multilayer perceptrons with feedback are studied and it is shown that their dynamics provides a built-in time-warping invariance, as required for presentation speed fluctuations in speech recognition.

International Journal of Neural SystemsMOTION DETECTION AND DIRECTION DETECTION IN LOCAL NEURAL NETS

4 Citations1989H.-U. Bauer, T. Geisel

A model for motion and direction detection of moving pulses whose performance is independent of pulse velocity, size and shape is presented and is found to be invariant in a window of pulse velocities an order of magnitude wide.