Reservoir computing with solitons

Abstract

Reservoir computing is a promising framework that facilitates the approach to physical neuromorphic hardware by enabling a given nonlinear physical system to act as a computing platform. In this work, we exploit this paradigm to propose a versatile and robust soliton-based computing system using a discrete soliton chain as a reservoir. By taking advantage of its tunable governing dynamics, we show that sufficiently strong nonlinear dynamics allows our soliton-based solution to perform accurate regression and classification tasks of non-linear separable datasets. At a conceptual level, the results presented pave a way for the physical realization of novel hardware solutions and have the potential to inspire future research on soliton-based computing using various physical platforms, leveraging its ubiquity across multiple fields of science, from nonlinear optical media to quantum systems.

Transmission of electromagnetic waves through nonlinear optical layers

“…In particular, we will show that, for soliton excitations to occur in layers with a de- focusing Kerr nonlinearity, the incident electromagneticwave intensity must exceed a specific minimum value. This finding opens up the possibility of predicting intensity regions where the formation of solitons would take place and other ones where such effects would not be observed at all, and could be of relevance in opticalcomputing investigations [10,11].…”

Reservoir Computing with Superconducting Electronics

“…It has been known for some time that these flux excitations are solitons: a consequence of the JTL's dynamics being governed by a perturbed Sine-Gordon equation in the discrete limit [14]. Recent work has explored the general RC capacity offered by soliton systems [15,16], and invites the question of whether the success of the JTL reservoir is shared with a broader class of systems. After all, the first implementation of a physical reservoir computer was based on complex wave interactions in shallow water [5], one of the canonical systems for the study of soliton dynamics [17,18].…”

Training Gaussian boson sampling by quantum machine learning

“…Unsupervised and supervised learning endow new designs for quantum circuits (Marquardt 2021), metrology and cryptography (Lumino et al 2018;Fratalocchi et al 2021), multilevel gates , and Bell tests (Melnikov et al 2020). NN are also triggering new fundamental investigations in quantum neuromorphic and wave computing Hughes et al 2019;Ballarini et al 2020;Nokkala et al 2020;Marković and Grollier 2020;Silva et al 2021), quantum thermodynamics (Sgroi et al 2021), and topological photonics (Pilozzi et al 2021).…”