Li Pengcheng, Wang Kesheng, Jiang Shanshan, He Gang, Zhang Hainan, Cheng Shuo, Li Qingxuan, Zhu Yixin, Fu Can, Wei Huanhuan, He Bo, Li Yujiao
School of Integrated Circuits, Anhui University, Hefei 230601, China.
School of Materials Science and Engineering, Anhui University, Hefei 230601, China.
Nanomaterials (Basel). 2024 Sep 29;14(19):1573. doi: 10.3390/nano14191573.
The traditional computer with von Neumann architecture has the characteristics of separate storage and computing units, which leads to sizeable time and energy consumption in the process of data transmission, which is also the famous "von Neumann storage wall" problem. Inspired by neural synapses, neuromorphic computing has emerged as a promising solution to address the von Neumann problem due to its excellent adaptive learning and parallel capabilities. Notably, in 2016, researchers integrated light into neuromorphic computing, which inspired the extensive exploration of optoelectronic and all-optical synaptic devices. These optical synaptic devices offer obvious advantages over traditional all-electric synaptic devices, including a wider bandwidth and lower latency. This review provides an overview of the research background on optoelectronic and all-optical devices, discusses their implementation principles in different scenarios, presents their application scenarios, and concludes with prospects for future developments.
具有冯·诺依曼架构的传统计算机具有存储单元和计算单元分离的特点，这导致在数据传输过程中会消耗大量时间和能量，这也就是著名的“冯·诺依曼存储墙”问题。受神经突触的启发，神经形态计算因其出色的自适应学习和并行能力，成为解决冯·诺依曼问题的一种很有前景的解决方案。值得注意的是，2016年，研究人员将光集成到神经形态计算中，这激发了对光电和全光突触器件的广泛探索。这些光学突触器件相对于传统的全电突触器件具有明显优势，包括更宽的带宽和更低的延迟。本文综述了光电和全光器件的研究背景，讨论了它们在不同场景下的实现原理，介绍了它们的应用场景，并对未来发展前景进行了总结。
