Regensburg 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 39: Machine Learning in Dynamics and Statistical Physics II
DY 39.1: Vortrag
Donnerstag, 20. März 2025, 15:00–15:15, H47
Fast and energy-efficient reservoir computing using a resonant-tunneling diode — •Osamah Sufyan1, Antonio Hurtado2, and Kathy Lüdge1 — 1Technische Universität Ilmenau, Institut für Physik, Weimarer Straße 25, 98693 Ilmenau, Germany — 2University of Strathclyde, Institute of Photonics, Glasgow, United Kingdom
Resonant-tunneling diodes (RTDs) have garnered significant attention as platforms for neuromorphic computing, owing to their fast operation and intricate nonlinear dynamics. Among the most hardware-friendly and energy-efficient paradigms in this domain is reservoir computing (RC), where the nonlinear dynamics of a physical system are leveraged to perform complex computational tasks.
In this work, we explore the use of a single RTD as a reservoir, employing time-multiplexing techniques for chaotic time-series prediction achieving similar performance to previous RC approaches [1]. Our findings highlight the relationship between the RTD’s distinct dynamical regimes and the reservoir’s performance in predicting future values of the Mackey-Glass and Lorenz system time series. Additionally, we investigate the RTD as an excitable system, demonstrating its potential for spiking neural network applications. We examine various data encoding and decoding strategies for spike-based operations, further underscoring the versatility of RTDs in neuromorphic computing.
[1] L. Jaurigue and K. Lüdge, Neuromorph. Comput. Eng., 4, 014001 (2024).
Keywords: reservoir computing; non-linear dynamics