Exploiting Bistability and Viscoelasticity in Reservoir Computing
Year: 2025
Type: article
Abstract: Reservoir computing (RC) is a powerful framework for processing time-domain signals, leveraging the nonlinear dynamics of physical systems to solve complex tasks. A key requirement for a nonlinear dynamical system to function effectively as a reservoir computer is its ability to retain recent inputs while gradually forgetting older ones. In electrical systems, this short-term memory is easily implemented using a delay line. However, in mechanical systems, viscoelasticity is the only mechanism for information retention, as the system’s state depends not only on its current deformation but also on its deformation history. Motivated by this, we study the efficiency of a viscoelastic bistable chain in RC, using a bistable oscillator coupled with a standard linear solid element as the fundamental unit cell. Effect of parameter variations in the context of information theory is studied, identifying regions of high performance. Notably, poor performance zones indicate that chaotic reservoir dynamics lead to deviations in long-term predictions. To further understand this, we apply Melnikov’s criterion to determine the chaotic threshold of the unit cells and pinpoint optimal RC performance regimes. Finally, the reservoir computing functionality of the bistable chain is demonstrated through an analog electrical implementation.
Source: Nonlinear Dynamics
Authors: Pritam Ghoshal, James M. Gibert, Anil K. Bajaj
Institution: Purdue University West Lafayette
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,790
Funder: Directorate for Engineering
Award: G950638150
OpenAlex ID: W4414261887