一类一般输入输出函数的离散神经元模型的分支

利用经典分支理论研究了一类一般输入输出函数的离散神经元模型的分支问题,得到了该类模型产生倍周期分支和鞍-结点分支的充分条件,推广了甘前特殊的正弦输入输出函数的该类模型的结果.所得的结果为这一类神经网络的应用提供了重要的理论基础.     

Rhythm Synchronization of Coupled Neurons with Temporal Coding Scheme

Encoding information by firing patterns is one of the basic neural functions, and synchronization is important collective behaviour of a group of coupled neurons. Taking account of two schemes for encoding information （that is, rate coding and temporal coding）, rhythm synchronization of coupled neurons is studied. There are two types of rhythm synchronization of neurons： spike and burst synchronizations. Firstly, it is shown that the spike synchronization is equivalent to the phase synchronization for coupled neurons. Secondly, the similarity function of the slow variables of neurons, which have relevant to the bursting process, is proposed to judge the burst synchronization. It is also found that the burst synchronization can be achieved more easily than the spike synchronization, whatever the firing patterns of the neurons are. Hence the temporal encoding scheme, which is closely related to both the spike and burst synchronizations, is more comprehensive than the rate coding scheme in essence.     

关于周期神经网络逼近阶的研究

借助于有关Fourier级数的Riesz平均构造出了一类含有一个隐含层的周期神经网络与平移网络,与已有的讨论相比较,在获得相同的逼近阶的情况下,此类网络的隐层单元要求较少的神经元个数.     

我的造梦工厂

正一大脑里面有很多脑神经细胞,每天都要帮助我们做各种各样的事情。二到了晚上,我们的身体累了,需要休息了。有些脑神经细胞却开始工作,它们干什么呢?造梦!瞧,下面,就是我的造梦工厂。     

面向神经形态感知和计算的柔性忆阻器基脉冲神经元

受人脑工作模式的启发,脉冲神经元作为人工感知系统和神经形态计算体系的基本计算单元发挥着重要作用.然而,基于传统互补金属氧化物半导体技术的神经元电路结构复杂,功耗高,且缺乏柔韧性,不利于大规模集成和与人体兼容的柔性感知系统的应用.本文制备的柔性忆阻器展示出了稳定的阈值转变特性和优异的机械弯折特性,其弯折半径可达1.5 mm,弯折次数可达10~4次.基于此器件构建的神经元电路实现了神经元的关键积分放电特性,且其频率-输入电压关系具有整流线性单元相似性,可实现基于转换法的脉冲神经网络中神经元的非线性处理功能.此外,基于电子传输机制和构建的核壳模型,对柔性忆阻器的工作机制进行分析,提出了电场和热激发主导的阈值转变机制;进一步对忆阻器和神经元的电学特性进行电路仿真模拟,验证了柔性忆阻器和神经元电路工作机制的合理性.本文对柔性神经元的研究可为神经形态感知和计算系统的构建提供硬件基础和理论指导.     

Complex Behavior in a Selective Aging Neuron Model Based on Small World Networks

Complex behavior in a selective aging simple neuron model based on small world networks is investigated. The basic elements of the model are endowed with the main features of a neuron function. The structure of the selective aging neuron model is discussed. We also give some properties of the new network and find that the neuron model displays a power-law behavior. If the brain network is small world-like network, the mean avalanche size is almost the same unless the aging parameter is big enough.     

电子直线加速器磁控管频率稳定的自适应线性神经元方法

电子直线加速器电子束能量的稳定与否取决于功率源工作频率的稳定性,磁控管在短时间内的散谱和轻微跳谱造成稳频系统的控制精度下降,最后电子束的扫描均匀度下降。引入自适应线性神经元方法(ADALINE)和噪声对消技术以消除对工作频率长期稳定性的影响,从而保证了电子束的扫描均匀度。     

经颅磁刺激感应外电场作用下最小神经元模型放电起始动态机理分析

经颅磁刺激(TMS)是一种采用电磁线圈在大脑指定区域产生磁场的刺激方式. TMS的治疗原理是通过电磁感应产生作用于神经元的外加电场进而影响神经元编码. 然而目前TMS感应外电场改变神经元编码的内在机理尚不清楚.研究表明, 神经元编码由神经元的放电起始动态机理决定. 本文建立了TMS感应外电场作用下的最小神经元模型, 采用相平面分析和分岔分析方法, 研究了外电场作用下神经元放电起始动态的动力学机理, 并从阈下电位的不同动力学特性离子电流竞争角度揭示了TMS感应外电 场作用下神经元放电起始动态的生理学机理.     

Complete Synchronization in Coupled Chaotic HR Neurons with Symmetric Coupling Schemes

Chaos synchronization in an array of coupled chaotic neurons with symmetric coupling is investigated. A criterion for the stability of the synchronization manifold is deduced by transforming the variational equation of the coupled system into a block diagonal one, and the critical coupling strengths for synchronization in different coupled cases are given. As examples for illustration, the HR neurons with the open-ended （i.e. chain）, ring, star and all-to-all （i.e. global） coupling schemes are considered. It is shown that the coupling scheme plays an important role in synchronization and information transmission of neurons.     

Complete Synchronization of Coupled Hindmarsh-Rose Neurons with Ring Structure

Complete synchronization of coupled Hindmarsh-Rose (HR) neurons with a ring structure is studied. The criterion for complete synchronization of coupled neurons with the ring structure is obtained through the stability analysis of the linearized synchronization error system. Numerical simulation is given to test the criterion for a system with three coupled HR neurons.