A Review of Recent Developments in Neuromorphic Computing Based on Emerging Memory Devices

Neuromorphic computing is a novel computing paradigm that mimics biological neural systems’ structure and information processing mechanisms. By leveraging highly parallel, low-power, brain-inspired architectures, neuromorphic computing provides efficient hardware support for artificial intelligence (AI). Within this framework, the synapse and neuron models and their circuit implementations are the foundational core of neuromorphic computing, significantly influencing its performance and capabilities. This paper reviews recent advances in neuron model circuits and the development of neuromorphic computing. Specifically, we discuss: 1) the basic working principles and implementation methods of synapses; 2) the biomimetic characteristics and physical circuits of various neuron models; 3) neural networks and corresponding dynamics analysis. Finally, future research directions for neuron model design and neuromorphic networks are discussed.