Four‐Channel Full‐Function Photonic Spiking Neural Network Chips for Gene Analysis

作者：Xingxing Guo, Ziwei Song, Shuiying Xiang, Haowen Zhao, Yahui Zhang, Yanan Han, Xinran Niu, Yizhi Wang, Wenzhuo Liu, Zhiquan Huang, Yue Hou, Yuechun Shi, Ye Tian, Yue Hao · 发表于：Laser & Photonics Review · 年份：2025 · DOI：10.1002/lpor.202500864 · 被引用次数：3 · 研究领域：Neural Networks and Reservoir Computing、Advanced Memory and Neural Computing、Photonic and Optical Devices

Abstract Brain‐inspired computing is essential for a range of critical computing tasks, including image processing, speech recognition, and applications of artificial intelligence and deep learning. However, compared to the real neural system, the traditional computing framework has a major limitation of physically separated storage and processing units, making it difficult to achieve fast, efficient, and low‐energy computing. To overcome this limitation, it is an attractive option to use hardware devices designed to simulate neurons and synapses. Once such hardware devices are integrated into neural networks or neuromorphic systems, their information‐processing methods will more closely resemble those of the human brain. Here, a four‐channel fully functional photonic spiking neural network architecture is proposed, in which a silicon photonic Mach‐Zehnder interferometer (MZI) network functions as the synapses performing the linear computation, and the Indium Phosphide (InP)‐based photonic integrated distributed feedback laser array with an intracavity saturable absorber (DFB‐SA) acts as the spiking neurons executing the nonlinear computation. In the experiment, through collaborative design of hardware algorithms, gene analysis tasks based on the HIV dataset and the Splice dataset are successfully completed with accuracy rates of 97.3% and 98%, respectively. The proposed hardware implementation of an all‐optical spiking neurosynaptic network architecture is expected to direct...