Nonlinear input transformations are ubiquitous in quantum reservoir computing. Issue 1 (1st March 2022)

• Record Type: Journal Article
• Authors: Govia, L C G; Ribeill, G J; Rowlands, G E; Ohki, T A
• Is Part Of: Neuromorphic computing and engineering. Volume 2:Issue 1(2022)
• Publication Date: 2022-03-01
• Subjects: quantum reservoir computing -- physical reservoir computing -- quantum neuromorphic computing; Neural networks (Computer science) -- Periodicals; Neural computers -- Periodicals; Neuromorphics -- Periodicals; 006.3
• DOI: 10.1088/2634-4386/ac4fcd
• Languages: English
• ISSNs: 2634-4386

Abstract

Abstract: The nascent computational paradigm of quantum reservoir computing presents an attractive use of near-term, noisy-intermediate-scale quantum processors. To understand the potential power and use cases of quantum reservoir computing, it is necessary to define a conceptual framework to separate its constituent components and determine their impacts on performance. In this manuscript, we utilize such a framework to isolate the input encoding component of contemporary quantum reservoir computing schemes. We find that across the majority of schemes the input encoding implements a nonlinear transformation on the input data. As nonlinearity is known to be a key computational resource in reservoir computing, this calls into question the necessity and function of further, post-input, processing. Our findings will impact the design of future quantum reservoirs, as well as the interpretation of results and fair comparison between proposed designs.