Answers are grounded in this paper and cite their sources.
Spiking neural networks (SNNs), inspired by the structure and function of the human brain, represent a promising approach to improving the energy efficiency of artificial intelligence (AI) computing. By employing massively parallel, event-driven processing, they overcome the inherent limitations of the conventional von Neumann architecture arising from serial processing and the memory wall problem. One of key requirements for implementing SNN hardware is the realization of artificial neurons, which generate spike signals once sufficient synaptic inputs are accumulated. This review summarizes and compares recent advances in single transistor neuron devices (1T-neurons), which offer excellent scalability and full compatibility with standard complementary metal-oxide-semiconductor (CMOS) technology. First, the working principle of 1T-neurons is primarily described, followed by an overview of their various structures and advanced functionalities. Next, cointegration strategies with artificial synaptic devices and sensory neurons for in-sensor SNNs are introduced. Finally, the challenges and prospects of developing 1T-neurons are discussed.
No takes yet. Share an insight, caveat, or question.
Yun et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: