SAMOP 2021 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
QI: Fachverband Quanteninformation
QI 7: Quantum Information: Applications
QI 7.4: Vortrag
Mittwoch, 22. September 2021, 11:30–11:45, H4
The Dicke Model as an Associative Quantum Neural Network — •Lukas Bödeker1, Eliana Fiorelli1,2, and Markus Müller1,2 — 1Institute for Quantum Information, RWTH Aachen University, D-52056 Aachen, Germany — 2Peter Gruenberg Institute, Theoretical Nanoelectronics, Forschungszentrum Juelich, D-52425 Juelich, Germany
Nowadays Classical Artificial Neural Networks (NNs) show their great power and versatility in information processing tasks. Early instances of NNs are given by Associative NNs, that have the ability to retrieve a stored state, starting from a compromised initial one. Such dynamics can be engineered via a stochastic evolution, where stored configurations are minima of an energy landscape. One of the first examples of associative NNs is the Hopfield NN, which is an Ising-type system featuring all-to-all interactions. Motivated by the fast progress in controlling quantum systems, as well as in quantum computation, a question that is currently explored is whether a Hopfield-type associative memory could be hosted in quantum systems. The goal is to understand whether quantum effects can be advantageous to store information. To this end, we consider the multi-mode Dicke model, in which a bosonic bath mediates an effective all-to-all spin interaction. The latter can be exploited to store information associatively, by setting the spin boson couplings accordingly. We analyse the storage properties of this system and further aim at investigating the maximum capacity i.e. the maximum number of stored states given a certain system size, generalising the classical approach introduced by Gardner.