Dresden 2026 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 6: Machine Learning in Dynamics and Statistical Physics I
DY 6.1: Vortrag
Montag, 9. März 2026, 09:30–09:45, HÜL/S186
Reservoir Computing with Hydrodynamically Coupled Active Colloidal Oscillators — •Veit-Lorenz Heuthe1, Lukas Seemann1, Samuel Tovey2, and Clemens Bechinger1, 3 — 1Universität Konstanz, Konstanz, Germany — 2Universität Stuttgart, Stuttgart, Germany — 3Centre for the Advanced Study of Collective Behavior, Konstanz, Germany
Reservoir computing is a newly emerging framework that exploites the dynamical response of complex physical systems to external perturbations. The high-dimensional, non-linear dynamics of active matter systems with hydrodynamic interactions offsers great potential for highly tunable physical reservoirs. Here, we present a physical reservoir that exploits the hydrodynamic interactions between several hundred colloidal oscillators for chaotic timeseries forecasting. We demonstrate that the inherent memory in this system facilitates detection of hidden anomalies in non-Markovian time-signals. Our results highlight the potential of active matter for locating subtle, non-disruptive signatures in e.g. financial stock markets, physiological measurements or seismic and climate data. Achieving such computing functinalities in physical systems could enable the development of intelligent hardware for edge-computing.
Keywords: Reservoir Computing; Active Colloidal Particles; Forecasting; Anomaly Prediction