Reservoir Computing with Random Magnetic Textures
In this talk we will discuss how a random magnetic “fabric” composed of skyrmion clusters can be effectively employed to implement a computational technique known as reservoir computing [1]. This is achieved by leveraging the nonlinear resistive response of individual skyrmions to driving currents arising from the anisotropic magneto-resistance effect (AMR). Time-varying current signals injected via contacts into the magnetic substrate are shown to be modulated by the fabric’s AMR due to the current distribution following paths of least resistance as it traverses the geometry. By tracking resistances across multiple input/output contacts, we show how the instantaneous current distribution effectively carries temporally correlated information about the injected signal. This in turn allows us to numerically demonstrate simple pattern recognition. We argue that the fundamental ingredients for such a device to work are threefold [2]: i) Concurrent probing of the magnetic state; ii) stable ground state; iii) nonlinear response to input forcing. Whereas this talk will only consider the use of skyrmion fabrics, the basic ingredients or reservoir computing will be argued to be general enough to spur the interest of the greater complex materials community to explore novel reservoir computing systems.
Authors : D. PINNA (1), G. BOURIANOFF (2) and K. EVERSCHOR-SITTE (1)
(1) Physics Department, Johannes Gutenberg University
(2) Intel Corp. (retired)
[1] G. Bourianoff, D. Pinna, M. Sitte, K. Everschor-Sitte, AIP Advances 8 (5), 055602 (2018)
[2] D. Pinna, G. Bourianoff, K. Everschor-Sitte, arXiv:1811.12623 (2019)
(*) Persons who do not have a CEA badge must request an entry authorization sufficiently in advance, before November 21th at sabrina.megias@cea.fr or celine.conche@cea.fr