光神经形态计算研究进展与展望(特邀)

脑科学与类脑研究是国际必争战略性前沿。人工智能与深度学习的飞速发展对算力提出了迫切需求。而传统的冯诺依曼架构,由于存算分离导致功耗墙和内存墙,摩尔定律也逐渐放缓。光神经拟态计算充分融合高速光通信、光互连、光集成、硅基光电子与神经拟态计算的特点,具有超高速、大带宽、多维度等优势,在高性能计算、人工智能领域有广阔的应用前景,是突破后摩尔时代传统微电子计算极限极具竞争力的方案。本文回顾了国内外主要研究团队在光神经元、光突触、光神经网络的理论、算法及器件方面的工作,并提出了展望。     

Synchronization transition of a coupled system composed of neurons with coexisting behaviors near a Hopf bifurcation

The coexistence of a resting condition and period-1 firing near a subcritical Hopf bifurcation point, lying between the monostable resting condition and period-1 firing, is often observed in neurons of the central nervous systems. Near such a bifurcation point in the Morris-Lecar （ML） model, the attraction domain of the resting condition decreases while that of the coexisting period-1 firing increases as the bifurcation parameter value increases. With the increase of the coupling strength, and parameter and initial value dependent synchronization transition processes from non-synchronization to compete synchronization are simulated in two coupled ML neurons with coexisting behaviors： one neuron chosen as the resting condition and the other the coexisting period-1 firing. The complete synchronization is either a resting condition or period-1 firing dependent on the initial values of period-1 firing when the bifurcation parameter value is small or middle and is period- 1 firing when the parameter value is large. As the bifurcation parameter value increases, the probability of the initial values of a period- 1 firing neuron that lead to complete synchronization of period- 1 firing increases, while that leading to complete synchronization of the resting condition decreases. It shows that the attraction domain of a coexisting behavior is larger, the probability of initial values leading to complete synchronization of this behavior is higher. The bifurcations of the coupled system are investigated and discussed. The results reveal the complex dynamics of synchronization behaviors of the coupled system composed of neurons with the coexisting resting condition and period-1 firing, and are helpful to further identify the dynamics of the spatiotemporal behaviors of the central nervous system.     

简谐噪声激励下FitzHugh-Nagumo神经元的动力学行为

研究了简谐噪声激励下的FitzHugh-Nagumo神经元模型, 其放电形式、相干共振等动力学行为均受噪声阻尼参数和频率参数的影响.选择不同的参数所得到的神经元的放电形式不同.神经元存在共振特性,对某一频率的噪声有更强的响应,在此频率参数下的峰序列更有序,出现相干共振系数的极小值.噪声的阻尼参数越大,不同的频率成分越多,神经元的响应也变得杂乱,进而导致神经元与噪声的同步变弱,峰序列相干共振系数也相应增大.     

Adaptive Regulation for Noise-Aided Weak Signal Detection in Spiking Neurons

Spiking neurons usually change their membrane properties, especially ion channel activity, during adaptation or synaptic modification to improve information processing and transmission. Using simple and biophysically realistic models, our analyses reveal that activity-dependent regulation of membrane properties contributes to sensitivity adaptation that improves the neuron ability of detecting sub-threshold signals in the presence of background noises. The improvement is achieved by regulating the conductance of ion channels on the membrane, dependent on the neuron firing activity.     

谷胱甘肽修饰的金纳米粒子探针比色法检测神经元素3

采用柠檬酸钠还原法合成了粒径为13 nm的金纳米粒子,并在其表面修饰上谷胱甘肽(GSHAuNPs)。在一定的盐浓度范围内,谷胱甘肽能保护金纳米粒子免受盐诱导的聚集。当神经元素3(Neurogenin3,ngn3)多肽片段存在时,在一定的盐浓度下,ngn3片段能够诱导GSH-AuNPs发生聚集,使金纳米粒子溶液由红色变成蓝色。以谷胱甘肽修饰的金纳米粒子为探针,建立了快速检测ngn3片段的比色方法。通过优化得到的最适实验条件为:ngn3与GSH-AuNPs的平衡反应时间10 min,缓冲液pH=6.0,NaCl浓度100 mmol/L。在优化条件下,检测ngn3的线性范围为20~300μg/L,检出限(LOD)为8μg/L。结果表明,本方法具有良好的选择性,可用于实际样品的检测。     

A Quantitative Analysis of Modulation of the Fast Excitatory Postsynaptic Potential of Neurons in Rat Sympathetic Ganglia by a Low-Intensity Laser

We propose a mathematical model which can quantitatively describe the kinetics of postsynaptic membrane potential modulated by a low intensity laser. By virtue of the numerical computation, it is found that the height of peak of fast excitatory postsynaptic potential （f-EPSP） for intermediate laser power density （around 2mW/cm^2） is higher than that for both small and large laser power densities, that is, of the f-EPSP peak height changes with the increasing laser power density （up to 5 m W/cm^2） in non-monotonic behaviour. The theoretical results are qualitatively consistent with the previous experimental data.     

Pattern Formation and Continuation in a Trineuron Ring with Delays

In this paper, we consider a single-directional ring of three neurons with delays. First, linear stability of the model is investigated by analyzing the associated characteristic transcendental equation. Next, we studied the local Hopf bifurcations and the spatio-temporal patterns of Hopf bifurcating periodic orbits. Basing on the normal form approach and the center manifold theory, we derive the formula for determining the properties of Hopf bifurcating periodic orbit, such as the direction of Hopf bifurcation. Finally, global existence conditions for Hopf bifurcating periodic orbits are derived by using degree theory methods.     

自然科学--请不要遗忘女性

科学与性别有关吗？初看上去，许多人对这个问题的回答似乎是否定的．自然科学是对世界的客观认识，与性别会有什么样的关系呢？然而，稍微深入地想一想，就会发现，其实科学与性别并非无关．人们经常谈论的一个问题就是：为什么杰出的女性自然科学家人数如此之少？以物理界为例，世界上法国女物理学家的比例最高；意大利其次；美国从事物理学研究的总人数也在下降，     

LONWORKS 网络技术在海军装备领域的应用前景

本文从介绍ＬＯＮＷＯＲＫＳ网络技术入手，论述了ＬＯＮＷＯＲＫＳ技术在海军装备上应用的广阔前景，提出了一条针对我国国情、军情，对改造已有装备以及建立新的系统都有着重要意义的新途径     

分解定理的网络结构分析

本文首先将逻辑神经元推广为模糊逻辑神经元，在此基础上证明了有关分解定理建模的两个重要定理，最后给出了分解定理的网络表示．