Spikoder: Dual‐Mode Graphene Neuron Circuit for Hardware Intelligence

Neuromorphic computing envisions the realization of a hardware neural network mimicking the brain's energy efficiency and rapid information processing. Leveraging the multifunctional capabilities of core neuromorphic building blocks offers an efficient approach to developing compact and intelligent hardware systems. This article demonstrates a novel technique to employ a graphene memristor‐based leaky integrate‐and‐fire circuit, which is hereby referred to as Spikoder. This circuit not only functions as a dynamic encoder transforming continuous input signals into spike sequences but also operates as a neuron circuit with reduced topological complexity. The circuit exhibits exceptional spike‐encoding performance, validated using spike‐encoded images from the Modified National Institute of Standards and Technology dataset. The effectiveness of this encoding technique is evaluated through experiments on single‐layer and double‐layer fully connected spiking neural networks (SNNs). The single‐layer SNN utilizing the dual‐mode neuron circuit achieves a high image recognition accuracy of 90.77%, while the implementation of a double‐layer SNN increases the test accuracy to 97.37%, further demonstrating its scalability for high‐level neuromorphic computing. This research highlights the possibility of using the hybrid encoder‐neuron circuit as an efficient and scalable solution for advanced neuromorphic computing hardware.