Regensburg 2025 – wissenschaftliches Programm

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Research with AI: Hardware, Software, Tools

AKPIK 3.2: Vortrag

Dienstag, 18. März 2025, 11:30–11:45, H5

Brain-inspired Computing with Gold Nanoparticle Networks: A Kinetic Monte Carlo Model — •Jonas Mensing and Andreas Heuer — Institute of physical Chemistry, University of Münster, Germany

Nanoparticles interconnected by insulating organic molecules exhibit nonlinear switching behavior at low temperatures. By assembling these nonlinear switches into a network and manipulating the inner charge transport dynamics via surrounding electrodes, the network can be functionalized to approximate functions such as Boolean Logic or model dynamical systems given the temporal dependence of input data. This makes nanoparticle networks promising candidates for neuromorphic computing and eventually bring machine learning applications on hardware.

We developed a kinetic Monte Calo simulation tool that applies established principles of single-electronics to model charge transport dynamics in nanoparticle networks. We demonstrate the network*s capability to approximate functions such as Boolean logic, perform nonlinear transformation of time dependent input signals, and forecast time series. These applications are evaluated using fitness metrics, enabling the optimization of surrounding electrode voltages to train the internal charge transport for a given task. The fitness measures are further analyzed in relation to system sizes, network disorder, or temporal scales. Furthermore, newly derived metrics enable us to link these design parameters to general nonlinear properties of the network.

Keywords: Neuromorphic Computing; Brain-inspired Computing; Nanoparticle; Reservoir Computing; Neural Networks