中心博士后马博文的工作——All-Optical Inhibitory Dynamics by Backscattering with Unidirectional Injection-Locked Photonic Spiking Neuron（单向注入锁定光子脉冲神经元后向散射的全光抑制动力学）的相关成果近期被IEEE Photonics Journal期刊接收发表，该工作得到了中国博士后科学基金（2023M732264）、国家自然科学基金（T2225023，62305213）的部分资助。
抑制动力学在生物尖峰神经元的信息处理中起着重要作用，其中传入的刺激削弱了神经元的活动水平。由于光子学中缺乏禁止物理效应，具有内在抑制动力学的光子尖峰神经元的实现以及有效抑制动力学在光学领域的应用一直是一个挑战。
针对光子神经学领域研究中抑制动力学的缺失，文章提出了一个具有全光抑制动力学的光子尖峰神经元，它是通过从单向发射的半导体环形激光器到主激光器的本征后向散射来实现的。文章进一步研究了半导体环形激光器在单向模式下的光子尖峰动力学，证明了光子尖峰神经元作为漏积分-放电神经元的计算特性。论文证明了在相同的输入信号下，随着后向散射强度的增加，产生的尖峰数减少。在后向散射的影响下，发射的尖峰数随着输入信号速率的增加而增加。在语音信号分类任务中，在抑制动态的影响下，分类准确率从89%提高到94%。这表明抑制动力学的存在有助于过滤掉不需要的干扰信息。结果表明，基于后向散射的抑制动力学在促进高级神经形态信息处理方面具有潜力。
摘要: Inhibitory dynamics plays an important role in the information processing of biological spiking neurons, where incoming stimuli weaken the activity level of neurons. Due to the lack of prohibitive physical effects in photonics, the realization of the photonic spiking neuron (PSN) with intrinsic inhibitory dynamics and the application of efficient inhibitory dynamics in the optical domain have been challenging. Herein, we propose a PSN with all-optical inhibitory dynamics, which is achieved by intrinsic backscattering from the unidirectional-emitted semi-conductor ring laser (SRL) to the master laser. The photonic spiking dynamics of the SRL operating in unidirectional mode were investigated, demonstrating the computational properties of the PSN as a leaky integrate-and-fire neuron. We prove that as the backscattering intensity increases, the number of spikes generated under the same input signal decreases. Under the influence of backscattering, the number of spikes fired increases with the input signal rate. In the speech signal classification task, the classification accuracy improves from 89% to 94% under the influence of inhibitory dynamics. This suggests that the presence of inhibitory dynamics helps filter out unwanted interference information. The results show the potential of backscattering-based inhibition dynamics for facilitating advanced neuromorphic information processing.