Autonomous Learning by Dynamical Systems with Inertial or Delayed Feedbacks

TL;DR

This paper explores how dynamical systems can autonomously adapt their behavior using inertial or delayed feedbacks, improving previous methods and extending applicability to systems without delays.

Contribution

It introduces an improved method for autonomous adaptation in dynamical systems, extending previous delayed feedback approaches to include inertial feedbacks and adaptable systems.

Findings

Effective adaptation demonstrated in oscillator networks

Method works with both delayed and inertial feedbacks

Numerical tests confirm robustness and versatility

Abstract

Dynamical systems can autonomously adapt their organization so that the required target dynamics is reproduced. In the previous Rapid Communication [Phys. Rev. E 90,030901(R) (2014)], it was shown how such systems can be designed using delayed feedbacks. Here, the proposed method is further analyzed and improved. Its extension to adaptable systems, where delays are absent and inertial feedbacks are instead employed, is suggested. Numerical tests for three different models, including networks of phase and amplitude oscillators, are performed.