Learning the Chaotic Hénon Map with a Multilayer Perceptron: Prediction Accuracy and Dynamical Analysis

This paper investigates the ability of a multilayer perceptron (MLP) to approximate the dynamics of the classical Hénon map. The reference system is analyzed through its fixed points, Jacobian matrix, local stability, phase portrait, bifurcation diagrams, and maximum Lyapunov exponent. The MLP is trained to learn the transition from consecutive states, ( x n , y n ) ↦ ( x n + 1 , y n + 1 ) , using two hidden layers and the Levenberg–Marquardt algorithm. The training results yield a minimum validation mean squared error of 2.1488 × 10 − 10 . Model performance is further assessed through free-running trajectories, phase-space representations, residual distributions, parameter-dependent bifurcation diagrams, and Lyapunov-exponent estimates. The results indicate qualitative similarities between the original system and the learned model. However, divergence in iterated trajectories highlights the limitations of one-step prediction accuracy for reproducing long-term behavior in chaotic systems. Parameter-dependent comparisons are interpreted cautiously because the role of the control parameter in the learned model is not fully specified.

Authors

Institutions

Publication Details

Journal
PROOF
Published
2026-11-17
DOI
https://doi.org/10.37394/232020.2026.6.6
Primary Topic
Chaos control and synchronization
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Learning the Chaotic Hénon Map with a Multilayer Perceptron: Prediction Accuracy and Dynamical Analysis

Hacene Gharout
1 links
PROOF
Chaos control and synchronization
article

Learning the Chaotic Hénon Map with a Multilayer Perceptron: Prediction Accuracy and Dynamical Analysis

Hacene Gharout
article en

Abstract

This paper investigates the ability of a multilayer perceptron (MLP) to approximate the dynamics of the classical Hénon map. The reference system is analyzed through its fixed points, Jacobian matrix, local stability, phase portrait, bifurcation diagrams, and maximum Lyapunov exponent. The MLP is trained to learn the transition from consecutive states, ( x n , y n ) ↦ ( x n + 1 , y n + 1 ) , using two hidden layers and the Levenberg–Marquardt algorithm. The training results yield a minimum validation mean squared error of 2.1488 × 10 − 10 . Model performance is further assessed through free-running trajectories, phase-space representations, residual distributions, parameter-dependent bifurcation diagrams, and Lyapunov-exponent estimates. The results indicate qualitative similarities between the original system and the learned model. However, divergence in iterated trajectories highlights the limitations of one-step prediction accuracy for reproducing long-term behavior in chaotic systems. Parameter-dependent comparisons are interpreted cautiously because the role of the control parameter in the learned model is not fully specified.

PROOFVol. 6
University of Béjaïa (DZ)
Chaos control and synchronization
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon