Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MA: Fachverband Magnetismus
MA 41: Poster III
MA 41.34: Poster
Donnerstag, 20. März 2025, 15:00–17:30, P3
Gesture recognition with Brownian reservoir computing using geometrically confined skyrmion dynamics — •Grischa Beneke1, Thomas Brian Winkler1, Klaus Raab1, Maarten A. Brems1, Fabian Kammerbauer1, Pascal Gerhards2, Klaus Knobloch2, Sachin Krishnia1, Johan Mentink3, and Mathias Kläui1,4 — 1Institut für Physik, Johannes Gutenberg-Universität Mainz, Germany — 2Infineon Technologies Dresden, Germany — 3Radboud University, Institute for Molecules and Materials, the Netherlands — 4Center for Quantum Spintronics, Norwegian University of Science and Technology, Norway
Physical reservoir computing utilizes complex physical systems' dynamics for efficient information processing, minimizing training and energy requirements. Magnetic skyrmions, topologically stabilized spin textures, offer promising reservoir computing capabilities through their stability, strong non-linear behaviour, and energy-efficient manipulation. We demonstrate a time-multiplexed skyrmion reservoir computing approach to overcome traditional limitations in temporal pattern recognition [1]. By aligning the reservoir's timescales with real-world data, our approach processes hand gestures captured by Range-Doppler radar. This method scales to the nanometer regime and demonstrates competitive or superior performance compared to energy-intensive software-based neural networks. Our hardware approach's key advantage is its ability to integrate sensor data in real-time without temporal rescaling, enabling numerous applications. [1] G. Beneke et al., Nat. Commun. 15, 8103 (2024).
Keywords: Skyrmions; Reservoir Computing; Unconventional computing; Brownian computing