Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MA: Fachverband Magnetismus

MA 42: Posters Magnetism I

MA 42.47: Poster

Mittwoch, 18. März 2020, 15:00–18:00, P3

Designing balanced magnetic logic gates by means of neural networks — •Lukas Holzbeck, Daniele Pinna, and Karin Everschor-Sitte — Institut für Physik, Johannes Gutenberg Universität Mainz, D-55099 Mainz, Germany

Replacing traditional charged-based logic by nanomagnetic logic promises in particular the advantage of non-volatility. However to not lead to erroneous results upon integration to circuits, balanced logic gates are required. So far the search for such logic gates was rather on a trial and error base [cite https://journals.aps.org/prapplied/pdf/10.1103/PhysRevApplied.9.034004]. While such a search is consistently possible with only a few nanomagnets, it becomes rather infeasible for arrangements involving more nanomagnets. Our goal is to employ machine learning to design magnetic logic gates according to developed metrics of well-balanced and logical consistency. We train a neural network on data from static self-interacting systems and aim at finding generalized rules how to construct larger balanced gates and more complex logic gate structures such as the half adder.