Robust Neuromorphic Computation Under Device Variability and Non-Idealities

Neuromorphic engineering draws inspiration from biological neural systems, which operate robustly despite significant variability, noise, and heterogeneity at the level of neurons and synapses. Understanding how neural systems maintain reliable function under such conditions is central to the development of energy-efficient, brain-inspired computing.As device scaling approaches physical limits, emerging non-volatile technologies - such as memristors and quantum-dot-based transistors - offer new opportunities for implementing neural computation through computing-in-memory (CIM) and adaptive neuromorphic architectures. However, these technologies introduce variability, endurance constraints, non-ideal switching dynamics, and stochastic behavior, which directly impact the implementation of neural computation in hardware.This Research Topic focuses on how device-level non-idealities affect and can be leveraged within neuromorphic systems that implement biologically inspired computation. We encourage contributions that connect device variability to neural coding, spiking dynamics, synaptic plasticity, and network-level function. Biological systems provide key inspiration, using redundancy, population coding, and adaptive learning to achieve robust computation under variability. Such systems also provide testbeds for investigating how biological neural circuits maintain reliable computation under variability and noise.We welcome interdisciplinary work linking device variability characterization with neuromorphic system design, including compact modeling, circuit and architecture co-design, and hardware-aware learning. Contributions must demonstrate how innovations enable scalable neuromorphic systems, rather than focusing solely on device physics or material properties.Submissions should demonstrate clear relevance to neuromorphic principles, such as spiking or event-driven computation, synaptic plasticity, and on-chip learning. Studies focusing solely on device physics, materials science, or standalone device modelling without integration into neuromorphic systems will be considered out of scope.Topics of interest include:• Device variability and stochastic effects in spiking and neuromorphic systems • Impact of non-idealities on neural coding (rate, temporal, population coding) • Robustness of event-driven and spike-based computation under variability • Hardware-aware learning in spiking neural networks (SNNs) • Synaptic plasticity mechanisms under non-ideal device conditions • In-situ and adaptive learning in hardware-constrained neuromorphic systems • Circuit and architecture co-design for robust neuromorphic computation • Compact models linking device behaviour to neuromorphic system dynamics and neural computation performance• Benchmarking neuromorphic systems under realistic non-idealities • Experimental demonstrations of neuromorphic systems with device variability Topic editors: Son Ngoc Truong, Kyeong-Sik Min Source: https://www.frontiersin.org/research-topics/80069/robust-neuromorphic-computation-under-device-variability-and-non-idealities