Karl Franzens University Graz | Graz University of Technology | |
Photonic and Mixed Ionic/Electronic Neuromorphic Device Concepts for Artificial Neural Networks Artificial neural networks (ANN), inspired by biological nervous systems, enable signal processing beyond the capabilities of classical von Neumann computer architectures. Through dynamically adapting the connectivity (synaptic weights) in individual devices and by applying learning algorithms ANNs can offer in memory and tensor computing capabilities. Yet, to fully unleash the potential of hardware ANNs there is still a need for neuromorphic device concepts, which properly emulate all necessary synaptic functions adequately and allow for an easy integration into large scale hardware ANNs. In this contribution we will demonstrate novel photonic, plasmonic and mixed ionic/electronic neuromorphic single device concepts. We will show that all presented devices concepts can be used to mimic fundamental functions of a synapse, such as long-term and short-term plasticity showing potentiation and depression, and spike-time or spike-rate dependent plasticity. Based on these achievements we demonstrate integration of single devices into ANNs for demonstrating AND- or OR logic gate operations and linear classification operations. |