Memory in Conductive Fabrics for Reservoir Computing
Christoph Walter Senn
Mass-spring networks are a popular choice to numerically simulate physical reservoir computing systems. As these kinds of networks are also widely used in the simulation of different types of fabric, in this paper, we evaluate the use of conductive fabrics as a reservoir. Stretching such a fabric changes its conductivity. Then, by measuring this change in resistance, we are able to exploit a piece of fabric as a computational resource by encoding input data into forces that pull on the cloth. Our numerical experiments support the use of conductive fabrics as a computational reservoir. Further, we estimated the memory capacity of a hypothetical cloth reservoir. Our results suggest that conductive fabrics have the potential to be exploited for computation and thus can serve as a viable reservoir for reservoir computing.
Keywords: Reservoir computing, machine learning, conductive fabric, numerical simulation
DOI: 10.32908/ijuc.v19.220124