Silicon carbide nanowire-based multifunctional and efficient visual synaptic devices for wireless transmission and neural network computing

With the rapid development of big data and the internet of things, the current computing paradigms based on traditional Von Neumann architecture have suffered from limited throughput and energy inefficiency. The memristor-based artificial neural network computing system could be regarded as a promising candidate to overcome this bottleneck. In this study, silicon carbide (SiC) nanowire (NW)-based optoelectronic memristors are successfully developed, which can realize complex brain-like features such as dendritic neuron and Pavlov's learning. On the basis of the visual function, perception, storage, and in situ computing functions integrated within optoelectronic memristors have been achieved. More importantly, benefiting from the excellent computing power of the SiC NW visual synapses, the constructed spike neural network is capable of implementing the identification of early lung cancer lesions. The accuracy rate of detection exceeds 90% with only a few iterations, indicating promising applications in the medical field with high efficiency and accuracy. The present study provides a promising path for developing and promoting SiC-based integrating perception-storage-computation artificial intelligence devices for neuromorphic computing technology.

Automated summary

In this study, silicon carbide (SiC) nanowire-based optoelectronic memristors are developed, which can achieve complex brain-like features and integrate visual function, perception, storage, and in situ computing. The constructed spike neural network using SiC NW visual synapses can accurately identify early lung cancer lesions with a detection accuracy rate exceeding 90%. This research provides a promising path for developing SiC-based integrating perception-storage-computation artificial intelligence devices for neuromorphic computing technology.