Reconstructing hidden attractors with reservoir computing

Hidden attractors are difficult to locate as their basins of attraction do not intersect with any small neighborhood of an equilibrium point. We examine the possibility of reconstructing the hidden chaotic attractor from time series via reservoir computing, where the reservoir is built by a network of driven Kuramoto oscillators. We compare all-to-all coupling, nearest-neighbour ring coupling, and a ring with one modified oscillator equation which ensures that any attractor will be hidden. The numerical results suggest that ring coupling supports accurate reconstruction over a wider range of coupling and forcing strengths than all-to-all coupling. Searches near reservoir equilibria nevertheless show why agreement in short-term prediction and long-term statistics does not establish preservation of hiddenness. Numerics for the modified ring show that it has the highest performance which coincides with our intuition that reconstructing the global dynamics would yield a more faithful attractor reconstruction. We also show that this construction precludes global forgetting of initial conditions for constant inputs, although local conditional contraction along a nonconstant signal remains possible. The distinction between reconstructing attractor dynamics and preserving basin structure provides a basis for evaluating and designing reservoir models of hidden attractors.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-17
DOI
https://doi.org/10.48550/arxiv.2611.22918
Primary Topic
Neural Networks and Reservoir Computing
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Reconstructing hidden attractors with reservoir computing

Emel Khan, Thomas Geert de Jong, Mohammad Zaman
arXiv (Cornell University)
Neural Networks and Reservoir Computing
preprint

Reconstructing hidden attractors with reservoir computing

Emel Khan, Thomas Geert de Jong, Mohammad Zaman
preprint en

Abstract

Hidden attractors are difficult to locate as their basins of attraction do not intersect with any small neighborhood of an equilibrium point. We examine the possibility of reconstructing the hidden chaotic attractor from time series via reservoir computing, where the reservoir is built by a network of driven Kuramoto oscillators. We compare all-to-all coupling, nearest-neighbour ring coupling, and a ring with one modified oscillator equation which ensures that any attractor will be hidden. The numerical results suggest that ring coupling supports accurate reconstruction over a wider range of coupling and forcing strengths than all-to-all coupling. Searches near reservoir equilibria nevertheless show why agreement in short-term prediction and long-term statistics does not establish preservation of hiddenness. Numerics for the modified ring show that it has the highest performance which coincides with our intuition that reconstructing the global dynamics would yield a more faithful attractor reconstruction. We also show that this construction precludes global forgetting of initial conditions for constant inputs, although local conditional contraction along a nonconstant signal remains possible. The distinction between reconstructing attractor dynamics and preserving basin structure provides a basis for evaluating and designing reservoir models of hidden attractors.

arXiv (Cornell University)
Neural Networks and Reservoir Computing
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