BPCPPDYSOE21 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 24: Dynamics and Statistical Physics - Open Session
DY 24.1: Vortrag
Dienstag, 23. März 2021, 11:00–11:20, DYb
Analysing and Optimizing Nonlinear Memory Capacity of Photonic Reservoir Computing — •Felix Köster1, Serhiy Yanchuk2, and Kathy Lüdge1 — 1Institut für Theoretische Physik, TU Berlin, Hardenbergstraße 36, 10623 Berlin — 2Institut für Mathematik, TU Berlin, Hardenbergstraße 36, 10623 Berlin
Reservoir computing is a neuromorphic inspired machine learning paradigm that utilizes the naturally occurring computational capabilites of dynamical systems. In this work, we investigate the linear and nonlinear memory capacity of a delay-based class-A and class-B-laser reservoir computer via eigenvalue analysis and numerical simulations. We show that these two quantities are deeply connected, and thus the reservoir computing performance is predictable by analyzing the eigenvalue spectrum. We introduce two new quantities to describe the influence of the eigenvalue spectrum on the reservoir computer performance. The insight won by the eigenvalue analysis yields understanding and thus helps applying better performing reservoir systems for a broader range of tasks.