一种减少ART1网络“祖母细胞”现象的解决方案”

ART1网络的输出层中的某个神经元损坏,会导致该神经元所代表类别的模式信息全部消失——产生“祖母细胞”现象,这是ART1网络的一个很大的缺陷。通过对ART1网络结构的改进,在总结构上增加记忆层,记忆层中神经元分记忆神经元和控制神经元;在输出层中新增了辅助神经元,将原有神经元称为主神经元;定义了各种神经元的工作方式;提出一种能够减少ART1网络“祖母细胞”现象的解决方案。     

一种新型联想记忆神经网络

提出了一种新型联想记忆神经网络,记忆容量接近2^N,完全消除了假模式,具有多模式回忆能力,同时具有较高的记忆效率;模式的各个分量分别存储在网络的N个连接中,这些连接构成一个通过网络中所有神经元的被称为“模式环”的单环回路;网络中处理和传递的信号为神经元编号组成的序列,神经元执行一组处理这种序列的符号和逻辑运算;连接由“连接状态”和“禁止路径”组成,前者用于存储信息,后者用于抑制网络中出现的假模式。     

随机延时Hodgkin-Huxley神经网络的同步与联想记忆

基于神经系统微观生理结构,给出具有空间分布随机延时的神经元间耦合,而这种随机延时描述峰电位从突触前神经元到突触后神经元在轴突上传播所需要的时间．记忆由时空发放的神经元集群表达．通过加入并改变噪声的强度,研究了噪声对Hodgkin—Huxley（HH）神经元网络系统联想记忆的作用,在噪声涨落的作用下,系统取得了对不完整输入的记忆恢复,得到与熟知的随机共振完全一致的结果．     

基于双重竞争共振机制的模糊ART神经网络

提出了一种基于双重竞争共振机制的模糊ART神经网络模型．该模型将输入节点的竞争共振机制引入到输出类别节点，采用输入节点和输出节点双重竞争共振机制，形成了一种新的模糊ART结构，解决了传统模糊ART网络记忆不稳定问题．将该模型应用于图像分割，解决了传统模糊ART网络图像分割结果随警戒参数的升高而出现的过度分割．实验结果表明，和原始模糊ART算法相比，新算法具有更好的分类识别性能，在飞机目标识别中平均识别率比原始算法提高3％～5％．     

基于竞争型神经网络的城市空气质量分析

该文利用竞争型神经网络来对空气质量进行分类。竞争型神经网络模拟生物神经网络中的神经元之间的兴奋、抑制与竞争的机制,进行网络的学习与训练。竞争型神经网络的输出层神经元能够通过竞争以确定胜者,胜者将指出哪一种原型模式最能代表输入模式。将不同空气质量等级下的各空气指标作为原型模式,通过输入样本模式,利用竞争网络的竞争特点得到胜者,以此得出空气质量等级。     

抑制性突触输入对神经元发放率响应的影响

文章基于整合发放神经元模型,研究抑制性突触输入和外部周期刺激对神经元发放率的影响。数值结果表明:在突出输入等于阈值时,适当频率的抑制性突触输入对神经元发放率具有易化作用,对神经元发放率峰值具有延迟作用;在突出输入小于阈值时,抑制性突出输入抑制神经元膜电位发放;在外部刺激作用下,神经元发放率幅值与外部刺激有关,刺激频率越高,发放率幅值越小,刺激强度越大,发放率幅值越大。     

三维连接系数矩阵的脉冲耦合神经网络彩色图像分割

提出三维连接系数矩阵的脉冲耦合神经网络(3D-PCNN)模型,将二维连接系数矩阵扩展成三维,值取空间欧氏距离的倒数,提出指数上升的动态阈值。利用神经元脉冲同步发放特性和自动波特性,直接分割彩色图像。结果表明,3D-PCNN算法与其他分割算法相比,运行时间减少了25%以上;该算法能够将不同区域信息以多层次彩色显示,改变RGB分量输入顺序时,同样可以分辨出更多的图像细节信息,分割精度高。     

海马结构中DG对CA3的信号传递作用

利用海马结构中的细胞通道模型,数值研究海马中CA3-DG网络的传递特性.首先分析外界刺激对锥体神经元放电节律的影响,放电节律经历周期峰放电,倍周期分岔通向混沌,激变周期3放电进而演化混沌,最后周期簇爆发的完整放电模式变化过程.然后通过突触连接模型,构造CA3-DG神经系统模型,分析了网络中各种神经元突触后电流总和的计算公式,突触后电流对神经元放电节律的影响以及簇爆发的产生机理,网络结构的强大编码能力揭示了CA3结构在海马信息传递中的特性.模型分析包含突触传递的时滞影响,模型结果与海马发放的实验现象相符合.     

具有认知判别功能的自联想记忆神经网络模型

提出一种具有认判别功能的自联想记忆神经网络模型，对一个新的输入模式，当网络达到平衡点时，通过一种差别函数可差别该模式是否是网络中已存贮的模式，如不是则对其进行学习；如是则该平衡状态即为其联想结果，给出了模型及其学习规则，分析了网络吸引子半径与有关参数的关系及模型对输入模式的筛选作用，最后给出了计算结果。     

小鼠海马神经网络对情景体验进行实时编码的功能单元的发现与鉴别

为了在神经网络层次上研究大脑对情景记忆信息实时编码的原理,我们创制了96通道微电极阵列系统．利用该系统,在小鼠经历惊吓事件过程中,对其大脑海马区多达260个神经元的放电活动进行了同步观察和记录．我们发现,惊吓情景刺激促发了小鼠海马CAI区神经元的放电模式发生改变,且这种改变与惊吓事件的特性及发生环境密切相关．运用统计学中的模式判别法。这些神经元的放电活动在低维空间形成了明显的集群式编码模式．进一步运用“移动窗口扫描”法,我们不仅能观察到神经网络实时编码的动态变化过程,而且能追踪惊吓事件之后记忆痕迹在神经元网络上的再现轨迹及次数．我们的分析表明：大脑海马CA1区神经网络的编码能力来自于由一系列神经元簇组成的功能性编码单元,神经元簇中的各神经元个体通过协同放电,从而克服了单一神经元放电的随机可变性,使对外界信息的实时编码获得稳定的高信噪比．这些功能性编码单元集群的激活形式能转换为具有实时性的一串串二进制数码,这一转换使不同动物大脑对各种行为事件获得了简明而通用的分类与编码．     

Firing patterns of long-term cultured neuronal network on multi-electrode array

Spontaneous neuronal activity plays an important role in the development and plasticity of brain. To explore the developmental changes in the firing pattern of the neuronal networks in vitro, the hippocampal neurons were cultured on the multi-microelectrode array dish for over 14 weeks and the spontaneous activity was recorded. The results showed that random firing was observed in the 1st week and transformed into synchronized activity after two weeks, then tightly synchronized activity appeared in week 2 to 7 and finally the activities transformed into the random firing pattern. These results suggested three stages in the long-term development of neuronal network in vitro: the stage for connection, the stage of synchronized activity and the mature stage. Synchronized firing shown by spontaneous activity was an important phenomenon in high density cultured neuronal network and transformed patterns during development.     

A study on discharge features of interneuron in the hippocampal network

The firing of neurons in the hippocampal network has a close relationship with human memory and learning. In this paper, a numerical simulation of interneurons in the hippocampal network has been operated. It analyzes the influence of external stimulation on firing rhythms. The diversity of firing pattern, especially the circle of unit firing pattern, is shown by ISI.     

一种局域连接前向人工神经网络结构的研究

双隐层标准前馈（ＢＰ）网络只要其隐层节点数足够多就能解决任何形式的分类问题．应用标准（ＢＰ）网络识别多模式类分类问题时存在以下缺陷：（１）对不同模式类均使用相同数目的隐层元；（２）增加新模式类后，网络要重新学习；（３）网络识别的机理研究困难．笔者提出了一种局域连接前馈神经网络（ＬＣＮＮ）结构，其隐层神经元与输出神经元之间为局域连接，学习算法与ＢＰ算法类似．ＬＣＮＮ具有以下特点：（１）便于自构网络结构，提高网络的推广能力；（２）便于提取各模式类的不变特性；（３）具有较强的记忆能力，便于实现追加学习．以五种海底沉积层介质类型的分类识别为例，分别利用标准前馈（ＢＰ）网络与ＬＣＮＮ网络进行分类识别，结果表明：ＬＣＮＮ便于自构网络结构，具有追加学习的能力．     

Attractor dynamics of network UP states in the neocortex

The cerebral cortex receives input from lower brain regions, and its function is traditionally considered to be processing that input through successive stages to reach an appropriate output. However, the cortical circuit contains many interconnections, including those feeding back from higher centres, and is continuously active even in the absence of sensory inputs. Such spontaneous firing has a structure that reflects the coordinated activity of specific groups of neurons. Moreover, the membrane potential of cortical neurons fluctuates spontaneously between a resting (DOWN) and a depolarized (UP) state, which may also be coordinated. The elevated firing rate in the UP state follows sensory stimulation and provides a substrate for persistent activity, a network state that might mediate working memory. Using two-photon calcium imaging, we reconstructed the dynamics of spontaneous activity of up to 1,400 neurons in slices of mouse visual cortex. Here we report the occurrence of synchronized UP state transitions ('cortical flashes') that occur in spatially organized ensembles involving small numbers of neurons. Because of their stereotyped spatiotemporal dynamics, we conclude that network UP states are circuit attractors--emergent features of feedback neural networks that could implement memory states or solutions to computational problems.     

On-line, voluntary control of human temporal lobe neurons

Daily life continually confronts us with an exuberance of external, sensory stimuli competing with a rich stream of internal deliberations, plans and ruminations. The brain must select one or more of these for further processing. How this competition is resolved across multiple sensory and cognitive regions is not known; nor is it clear how internal thoughts and attention regulate this competition. Recording from single neurons in patients implanted with intracranial electrodes for clinical reasons, here we demonstrate that humans can regulate the activity of their neurons in the medial temporal lobe (MTL) to alter the outcome of the contest between external images and their internal representation. Subjects looked at a hybrid superposition of two images representing familiar individuals, landmarks, objects or animals and had to enhance one image at the expense of the other, competing one. Simultaneously, the spiking activity of their MTL neurons in different subregions and hemispheres was decoded in real time to control the content of the hybrid. Subjects reliably regulated, often on the first trial, the firing rate of their neurons, increasing the rate of some while simultaneously decreasing the rate of others. They did so by focusing onto one image, which gradually became clearer on the computer screen in front of their eyes, and thereby overriding sensory input. On the basis of the firing of these MTL neurons, the dynamics of the competition between visual images in the subject's mind was visualized on an external display.     

Inhibitory feedback required for network oscillatory responses to communication but not prey stimuli

Stimulus-induced oscillations occur in visual, olfactory and somatosensory systems. Several experimental and theoretical studies have shown how such oscillations can be generated by inhibitory connections between neurons. But the effects of realistic spatiotemporal sensory input on oscillatory network dynamics and the overall functional roles of such oscillations in sensory processing are poorly understood. Weakly electric fish must detect electric field modulations produced by both prey (spatially localized) and communication (spatially diffuse) signals. Here we show, through in vivo recordings, that sensory pyramidal neurons in these animals produce an oscillatory response to communication-like stimuli, but not to prey-like stimuli. On the basis of well-characterized circuitry, we construct a network model of pyramidal neurons that predicts that diffuse delayed inhibitory feedback is required to achieve oscillatory behaviour only in response to communication-like stimuli. This prediction is experimentally verified by reversible blockade of feedback inhibition that removes oscillatory behaviour in the presence of communication-like stimuli. Our results show that a sensory system can use inhibitory feedback as a mechanism to 'toggle' between oscillatory and non-oscillatory firing states, each associated with a naturalistic stimulus.     

基于竞争的联想存储器学习算法

提出两种基于竞争的神经网络联想存储器学习算法—CC算法和ACC算法 ,并证明算法得到的神经网络对任一输入模式的竞争收敛性 ,由CC算法得到的网络 ,利用 p n个神经元存储p个n维样本模式 ;每个样本点都是吸引中心 ,不存在假吸引中心 ;对任一输入模式 ,总被吸引到与之海明距离最小的样本点上 ;不产生拒识点 .ACC算法是CC算法的改进形式 ,所得网络可在自适应学习中收敛 ,竞争次数较CC算法大大降低 本文算法得到的网络在存储容量、容错能力方面好于Hopfield联想存储器及作为联想存储器使用的BP网络 .     

Small modulation of ongoing cortical dynamics by sensory input during natural vision

During vision, it is believed that neural activity in the primary visual cortex is predominantly driven by sensory input from the environment. However, visual cortical neurons respond to repeated presentations of the same stimulus with a high degree of variability. Although this variability has been considered to be noise owing to random spontaneous activity within the cortex, recent studies show that spontaneous activity has a highly coherent spatio-temporal structure. This raises the possibility that the pattern of this spontaneous activity may shape neural responses during natural viewing conditions to a larger extent than previously thought. Here, we examine the relationship between spontaneous activity and the response of primary visual cortical neurons to dynamic natural-scene and random-noise film images in awake, freely viewing ferrets from the time of eye opening to maturity. The correspondence between evoked neural activity and the structure of the input signal was weak in young animals, but systematically improved with age. This improvement was linked to a shift in the dynamics of spontaneous activity. At all ages including the mature animal, correlations in spontaneous neural firing were only slightly modified by visual stimulation, irrespective of the sensory input. These results suggest that in both the developing and mature visual cortex, sensory evoked neural activity represents the modulation and triggering of ongoing circuit dynamics by input signals, rather than directly reflecting the structure of the input signal itself.