Grid-graph modeling of emergent neuromorphic dynamics and heterosynaptic plasticity in memristive nanonetworks.

Abstract Self-assembled memristive nanonetworks composed of many interacting nano objects have been recently exploited for neuromorphic-type data processing and for the implementation of unconventional computing paradigms, such as reservoir computing. In these networks, information processing and computing tasks are performed by exploiting the emergent network behaviour without the need of fine tuning its components. Here, we propose grid-graph modelling of the emergent behaviour of memristive nanonetworks, where the memristive behaviour is decoupled from the particular and detailed behaviour of each network element. In this model, the memristive behavior of each edge is regulated by an analytical potentiation-depression rate balance equation deduced from physical arguments. By comparing modelling and experimental results obtained on nanonetworks based on Ag NWs, the model is shown to be able to emulate the main features of the emergent memristive behaviour and spatio-temporal dynamics of the nanonetwork, including short-term plasticity, paired-pulse facilitation and heterosynaptic plasticity. These results show that the model represents a versatile platform for exploring the implementation of unconventional computing paradigms in nanonetworks.