Rodrigo Martínez (University of the Balearic Islands, Spain)

Closed quantum systems may exhibit different dynamical regimes, such as Many-Body Localization or thermalization, that can affect their ability to process information. Specifically, we establish the role of dynamical phases of Ising spin networks in the field of quantum reservoir computing.

Reservoir computing is an unconventional computing paradigm that consists in exploiting classical or quantum dynamical systems to solve nonlinear and temporal tasks. We observe that the thermal phase of the spin model is naturally adapted to the requirements of reservoir computing while the localized phase is detrimental for the purposes of this computational approach, finding an improved performance for linear and mildly nonlinear tasks in the transition regime. We uncover the physical mechanisms behind optimal information processing capabilities of the spin networks, essential for future experimental implementations.

Link to the teams group “naXys Seminars” (unamur members) here

Link to the seminar here