Command neurons in Pleurobranchaea receive synaptic feedback from the motor network they excite

Command neurons that cause rhythmic feeding behavior in the marine mollusc Pleurobranchaea californica have been identified in the cerebropleural ganglion (brain). Intracellular stimulation of single command neurons in isolated nervous systems, semi-intact prepartions, and restrained whole animals causes the same rhythmic motor output pattern as occurs during feeding. During this motor output pattern, action potentials recorded intracellularly from the command neurons occur in cyclic bursts that are phase-locked with the feeding rhythm. This modulation results from repetitive, alternating bursts of excitatory and inhibitory postsynaptic potentials, which are caused at least in part by synaptic feedback to the command neurons from identified classes of neurons in the feeding network. Central feedback to command neurons from the motor network they excite provides a possible general physiological mechanism for the sustained oscillation of neural networks controlling cyclic behavior.     

基于多元特征融合技术的农村家居安防系统设计

采用EC5-1719CLDNA嵌入式平台作为系统中央服务器,GSM和GPRS网络技术实现报警和远程监测与遥控,ZigBee无线传感技术进行实时传感器信息采集和传送,并结合流式媒体技术、多元特征融合(Multicharacters fusion)技术,无线摄像头设计技术等先进技术对智能家居安防系统进行总体架构。特别是基于多元特征融合技术理念,采用神经网络融合策略,设计了准确率高、稳健性好、性能优越的门禁控制系统,极大地提升了农村家居安防系统的安防功能,整个系统实现了家居的安全防范和远程智能控制。     

基于神经元晶体管和忆阻器的 Hopfield 神经网络及其在联想记忆中的应用

随着人工智能的高速发展,越来越多的神经元模型相继被提出,现有的神经元电路主要由普通晶体管、运算放大器等高功耗器件构成,存在结构复杂、集成度不高、兼容性差、功耗高、阈值调节难度高的缺点.针对以上不足,首次提出了一种全新的神经元结构,该结构仅由神经元晶体管、忆阻器和普通电阻构成,相比传统神经元电路,不包含复杂的差分运算电路以及电流与电压信号的转换电路,电路结构简单,同时具有良好的电路兼容性,可用于大规模集成.该结构利用神经元晶体管的加权求和特性以及阈值可控功能来模拟神经元信息传导过程,同时利用阈值忆阻器的阈值特性和阻值连续变化能力来设定和更新突触权值,使得该新型神经元结构不仅能实现传统神经元电路功能的同时,还具有能耗低、阈值动态可控、权值可编程的优点,不仅极大地简化了网络结构,还能加强网络性能.其次,还提出了基于这种新型神经元结构的忆阻离散Hopfield神经网络,该忆阻神经网络有助于促进人工神经形态系统的硬件实现,使神经网络系统能耗降低,集成度极大地提高,将这种网络运用在联想记忆和彩色数字图像恢复中,进一步说明了基于全新神经元结构的忆阻离散hopfield神经网络的实用性以及有效性.     

基于遗传算法的BP网络在小流域侵蚀量预测中的应用

人工神经网络可用于流域水土流失的预测.针对BP神经网络收敛速度慢及容易陷入局部最优解的缺点,设计了基于遗传算法(GA)的优化BP神经网络.利用遗传算法特有优势,为BP网络的初始权值和阈值搜索全局最优解空间,经过BP算法迭代训练,进行预测.依据黄土高原沟壑区杨家沟小流域多年径流与泥沙的实测数据,对创建的侵蚀量模型进行训练和预测,取得了较高的预测精度和较快的收敛速度.     

Characterizing a mammalian circannual pacemaker

Many species express endogenous cycles in physiology and behavior that allow anticipation of the seasons. The anatomical and cellular bases of these circannual rhythms have not been defined. Here, we provide strong evidence using an in vivo Soay sheep model that the circannual regulation of prolactin secretion, and its associated biology, derive from a pituitary-based timing mechanism. Circannual rhythm generation is seen as the product of the interaction between melatonin-regulated timer cells and adjacent prolactin-secreting cells, which together function as an intrapituitary "pacemaker-slave" timer system. These new insights open the way for a molecular analysis of long-term timing mechanisms.     

葡萄病害智能诊断网络模型的比较

以选择构建葡萄病害智能诊断系统的最适算法为目的,考查了3种神经网络模型：BP、RBF、GRNN,并以12种葡萄主要病害为研究样本,用3种网络模型分别构建了诊断系统,然后通过样本进行训练与仿真,比较各网络模型的稳定性和准确度。提出以GRNN作为病害诊断的最适模型,其诊断准确率为96％,为整个辅助决策系统的开发奠定了良好的基础。     

神经网络技术在节水灌溉专家系统中的应用

针对传统专家系统开发和应用过程中存在的问题,利用神经网络的学习功能、记忆功能和并行处理的优势,提出了用神经网络建立专家系统的方法,构造了神经网络专家系统的基本框架,并介绍了神经网络专家系统在节水灌溉上的应用。     

遗传算法与模拟退火算法在神经网络优化中的性能分析

神经网络有以任意精度逼近未知函数的能力，所以被广泛应用于各种领域中。目前广泛应用于神经网络优化的方法是反向传播（BackPropagation，BP），但是BP的全局搜索能力很有限，而全局搜索方法是神经网络优化问题很有潜力的办法。文中研究了两种全局优化算法：遗传算法（GeneticAlgorithm，GA）和模拟退炎（Simu-lated Annealing，SA），并且比较了它们在神经网络优化中的性能。     

马尾松毛虫幼虫发生严重程度的预测研究

为了提高马尾松毛虫幼虫发生严重程度的预测精度,寻求简便准确的预测方法,采用时间平稳序列法、回归预测法、马尔科夫链法、BP神经网络法和列联表多因子多级相关分析法对安徽省潜山县1983—2014年的马尾松毛虫越冬代、一代和二代幼虫发生的严重程度进行预测,研究历史符合率,并用2015年和2016年的实际发生情况验证。结果表明,平稳时间序列法,列联表多因子多级相关分析法计算简便,预测结果准确;BP神经网络法和马尔科夫链法预测结果非常准确。回归模型中以当代卵盛期卵量预测当代幼虫发生严重程度的一元回归模型的预测结果准确性高,其余一元回归模型预测结果稍差,多元回归模型和逐步回归模型优于一元回归模型。BP神经网络模型是一种理想的预测模型。     

Model of the origin of rhythmic population oscillations in the hippocampal slice

One goal of mammalian neurobiology is to understand the generation of neuronal activity in large networks. Conceptual schemes have been based on either the properties of single cells or of individual synapses. For instance, the intrinsic oscillatory properties of individual thalamic neurons are thought to underlie thalamic spindle rhythms. This issue has been pursued with a computer model of the CA3 region of the hippocampus that is based on known cellular and synaptic properties. Over a wide range of parameters, this model generates a rhythmic activity at a frequency faster than the firing of individual cells. During each rhythmic event, a few cells fire while most other cells receive synchronous synaptic inputs. This activity resembles the hippocampal theta rhythm as well as synchronized synaptic events observed in vitro. The amplitude and frequency of this emergent rhythmic activity depend on intrinsic cellular properties and the connectivity and strength of both excitatory and inhibitory synapses.     

温室温度控制系统神经网络PID控制算法研究

在以优质、高效、高产为目的的现代化农业发展新阶段,温室自动化技术的研究受到广泛重视.对于温室自动控制系统,由于其非线性、强耦合、纯滞后、大惯性的自身特性,传统PID控制已难以满足高品质温室控制系统的需求.由于BP神经网络具有强大的学习能力及非线性映射性,将BP神经网络控制引入常规PID控制中,采用BP神经网络PID控制方案,设计温室温度的自动控制系统并进行仿真验证.仿真结果表明,相比于传统的PID控制系统,所设计的基于BP神经网络PID控制系统具有更强的自适应能力与稳健性,控制品质具有明显优势.     

Neuronal generation of the leech swimming movement

The swimming movement of the leech is produced by an ensemble of bilaterally symmetric, rhythmically active pairs of motor neurons present in each segmental ganglion of the ventral nerve cord. These motor neurons innervate the longitudinal muscles in dorsal or ventral sectors of the segmental body wall. Their duty cycles are phase-locked in a manner such that the dorsal and ventral body wall sectors of any given segment undergo an antiphasic contractile rhythm and that the contractile rhythms of different segments form a rostrocaudal phase progression. This activity rhythm is imposed on the motor neurons by a central swim oscillator, of which four bilaterally symmetric pairs of interneurons present in each segmental ganglion appear to constitute the major component. These interneurons are linked intra- and intersegmentally via inhibitory connections to form a segmentally iterated and inter-segmentally concatenated cyclic neuronal network. The network appears to owe its oscillatory activity pattern to the mechanism of recurrent cyclic inhibition.     

基于神经网络逆复合控制器的MDF施胶系统的控制设计

根据神经网络逆控制原理,提出了一种新的中密度纤维板(MDF)施胶系统控制器设计方法。将MDF施胶系统作为被控对象,利用静态神经网络和若干积分器组成的动态神经网络构造出被控系统的逆系统,该逆系统与被控系统串接在一起所组成的系统是伪线性的。通过将PID控制器与神经网络逆系统结合在一起,构建出神经网络逆复合控制器。仿真结果表明:神经网络逆复合控制器能使系统获得优良的跟踪性和抗干扰能力,可以很好地实现对MDF施胶系统的控制目的。     

地区现代化水平BP人工神经网络评价方法初探

本文首先分析了现代化的动态发展的内涵,再结合相关理论、原则建立评价指标体系。之后选取31个地区作为样本对建立的BP人工神经网络进行训练。最后以安徽省为例进行实证研究,以期待对我国地区现代化评价提供有益的参考。     

Facilitation of spindle-burst sleep by conditioning of electroencephalographic activity while awake

A slow-wave electroencephalographic rhythm recorded from the sensorimotor cortex of the waking cat has been correlated behaviorally with the suppression of movement. Facilitation of this rhythm through conditioning selectively enhances a similar pattern recorded during sleep, the familiar spindle burst. The training also produced longer epochs of undisturbed sleep. The specific neural mechanism manipulated during wakefulness appears to function also in sleep and to be involved with the regulation of phasic motor behavior.     

基于神经网络PID的纤维板调施胶控制系统的设计

基于神经网络的优点,设计了神经网络作为主控制器的对系统进行辨识的纤维板调施胶控制系统和神经网络作为辅助控制器来修正PID控制器的纤维板调施胶控制系统.分析了两种神经网络控制器的性能优劣.仿真表明,神经网络作为辅助控制器调节PID比神经网络作为主控制器的超调小,系统稳定时间更短.     

径向基函数人工神经网络在棉花耗水量预测中的应用

采用径向基函数人工神经网络的方法,利用MATLAB工具箱,以气象资料中的平均气温、日照时数、平均风速作为输入变量,建立了预测新疆石河子地区棉花耗水量的RBF神经网络预测系统,并通过实测数据的检验得出此预测系统网络模型的绝对误差最大为0．0962mm／d,最小为0．0008mm／d,平均为0．0519mm／d;相对误差最大为1．7995％,最小为0．0165％,平均为0．9664％,可见网络模型预测的准确度较高,较以往的线性模型更合理,并且此网络训练花费的时间短,仅需0．0780s。     

Generation of torsional and vertical eye position signals by the interstitial nucleus of Cajal

The neural integrator, which converts eye velocity signals into position signals, is central to oculomotor theory. Similar integrators are probably necessary in any neural system that changes and maintains muscular tension. The integrator for horizontal eye position is in the pons, but the locations of the vertical and torsional integrators have not been clearly defined. Recording three-dimensional eye movements in alert monkeys during microstimulation and pharmacological inactivation of midbrain sites showed that the interstitial nucleus of Cajal generates both the torsional and vertical eye position signals. Up and down signals are linked with clockwise signals in the right brain and counterclockwise signals in the left brain. This three-dimensional coordinate system achieves orthogonality and bilateral symmetry without redundancy and optimizes energy efficiency for horizontal visual scanning.     

BP神经网络在水稻病虫害诊断中的应用研究

介绍了BP神经网络算法及编码的内涵,探讨了基于BP神经网络的水稻病虫害专家系统的构建。研究表明,BP神经网络在水稻病虫害诊断中具有较强的应用价值。     

Neurobiological bases of rhythmic motor acts in vertebrates

The general principles governing the nervous control of innate motor acts in vertebrates are discussed. Particular consideration is given to the control of locomotion in both mammals and lower vertebrates. One in vitro model of the lamprey central nervous system has been developed. It can be maintained in vitro for several days and the motor pattern underlying locomotion can be elicited in isolated sections of the spinal cord. These findings now allow a detailed analysis of the underlying neural mechanisms. The hypothesis that different parts of the network controlling locomotion can be used in a variety of other motor acts, including learned ones, is reviewed.