Classification of hyperchaotic, chaotic, and regular signals using single nonlinear node delay-based reservoir...

Area of Science:

• Complex Systems Dynamics
• Machine Learning for Dynamical Systems

Background:

• Traditional Lyapunov exponent methods require accurate system models for classifying dynamics.
• Many real-world systems lack precise models, hindering dynamic state classification.
• A data-driven approach is crucial for analyzing complex, unmodeled systems.

Purpose of the Study:

• To introduce a method for classifying hyperchaotic, chaotic, and regular dynamics using only observational data.
• To demonstrate the effectiveness of delay-based reservoir computers for dynamic state separation.
• To extend machine learning applications in chaos detection, including hyperchaotic signals.

Main Methods:

• Utilizing single nonlinear node delay-based reservoir computers.
• Training reservoir computers on observational data from theoretical or experimental systems.
• Employing the Mackey-Glass and optoelectronic oscillator systems as testbeds.
• Testing the generalization capability of trained models on different systems.

Main Results:

• Achieved high classification accuracy: up to 99.61% for Mackey-Glass and 99.03% for optoelectronic oscillator systems.
• Demonstrated robust separation of hyperchaotic, chaotic, and regular dynamics.
• Showcased the ability of reservoir computers to classify the dynamics of unseen systems (e.g., sine-logistic modulation map).

Conclusions:

• Delay-based reservoir computers offer a powerful, model-independent tool for classifying dynamical states.
• This approach significantly advances chaos detection by enabling hyperchaos identification.
• The method provides a robust and accurate solution for analyzing complex dynamics from observational data.