Memristive Cellular Nonlinear Networks (Mem²CNN)
Design and experimental validation of memristive CNN arrays, funded by the German Research Foundation (DFG) within SPP 2262.
Cellular Nonlinear Networks (CNNs) are massively parallel analog computing architectures in which locally connected cells interact through resistive or reactive couplings. The Mem²CNN project investigates the integration of memristive devices as programmable synaptic elements within CNN arrays, enabling compact, low-power implementations of spatiotemporal signal processing tasks directly in hardware.
The project was funded by the German Research Foundation (DFG) under the Priority Programme SPP 2262 – “Memristive Devices Toward Smart Technical Systems” (memristec.de), and was carried out at the Chair of Fundamentals of Electrical Engineering, TU Dresden (Prof. Ronald Tetzlaff). Key contributions include variability-aware design methodologies that exploit rather than suppress the inherent stochasticity of resistive switching devices, as well as circuit-level models enabling large-scale simulation of memristive CNN arrays.
Experimental validation on fabricated memristive test structures demonstrated the feasibility of analog in-memory computation with CNN topologies, establishing a pathway toward energy-efficient edge processing hardware. The results contribute to the broader goal of co-designing memristive devices and computing architectures for real-world smart technical systems.