利用输入信号先验知识构造某些分类神经网络的研究

提出一种前向神经网络只有一个隐元就可以解决奇偶校验或对称性校验问题，讨论了如何根据输入模式的先验知识来推导隐元的传递函数的构造该分类神经网络的方法，和已有的研究工作相比，提出的前向神经网络是解决某些线性不可分的模式识别问题的最简单的网络结构，所提出的分析方法也可推广用于类似的网络构造问题。     

神经网络的本质逼近阶

运用多元函数逼近工具, 对三层前向人工神经网络逼近连续和可积函数的本质逼近阶进行了定量研究. 证明了当激活函数满足一定条件时, 对任意的连续或可积函数, 能具体构造有明确隐层单元下界的三层网络使之对被逼近函数任意逼近. 给出该类神经网络逼近的上、下界估计和本质逼近阶估计, 刻画所构造网络的逼近性能与网络隐层拓扑结构之间的关系. 特别地, 当被逼近函数为二阶Lipschitz函数时, 所建立的神经网络其逼近速度完全取决于被逼近函数的光滑性. 所获结果对逼近连续或可积函数类的前向神经网络具体构造及逼近能力刻画有重要的理论指导意义.     

近似指数型神经网络的本质逼近阶

针对一类比Sigmoid更为宽泛的指数型激活函数,证明了三层前向神经网络的本质逼近阶.特别地证明了对于定义在Rd中紧子集上的任意连续函数f,存在隐层单元数为m(n)=Bdm(fi,nn)<ε(n 1)d(其中222(,)11,1,Bdfn=2 π2dωfn 2ω2(f,·)为f的二阶连续模,n为不小于1/ε的任意正整数)的近似指数型神经网络Rnσ(d)使其逼近f的精度与速度满足222(,())11,1.d∞fRnσd≤2 π2dωfn 2同时,当f属于α-Lipschtz函数类时,网络达到其本质逼近阶n?α(0<α≤2),所获结果较完整地刻画了该类神经网络的逼近特征,并揭示了该类神经网络逼近性态与网络拓扑之间的相依关系.     

输出-阈值耦合神经网络及基于此的最短路问题求解

通过建立神经元输出和与其相邻神经元阈值之间的相互耦合, 提出了一种输出-阈值耦合神经网络, 实现了对在通常脉冲耦合神经网络中所存在的自动波现象的模拟, 并基于它的自动波现象, 提出了一种用输出-阈值耦合神经网络求解最短路问题的方法, 该方法具有所需神经元数目少、神经元和网络的结构简单、大规模并行计算等特点, 可用于求解非对称赋权图单起点多终点的最短路问题, 其所需的计算量(迭代次数)仅正比于最短路的长度, 而与图的复杂程度、所存在的通路总数等无关. 最后给出了最短路求解的例子.     

基于Bagging神经网络集成的燃料电池性能预测方法

利用三维质子交换膜燃料电池数学模型模拟研究了电池流道进、出口高度对电池性能的影响,然后将数值模拟结果作为神经网络模型的训练数据.以流道进、出口高度和电池电压值作为输入变量,以电池电流密度作为输出变量,建立了3层反向传播神经网络模型;然后利用Bagging集成学习方法对神经网络模型进行集成,构建了燃料电池性能预测方法.研究发现:与单一神经网络模型相比, Bagging神经网络集成模型预测精度更高,且所需模型训练数据量更少.此外对于超出训练数据以外的情形, Bagging神经网络集成模型仍然能够准确地预测燃料电池的性能,且精度良好,表明Bagging神经网络集成模型的鲁棒性较好,可用于更宽工况范围内燃料电池性能的快速预测.     

会计信息可比性对企业资源配置效率的影响

以A股上市公司2007～ 2011年的数据为样本,检验了会计信息可比性对企业资源配置效率的影响.研究发现:会计信息可比性与企业投资不足显著负相关,说明增强会计信息可比性能够促进企业增加投资,从而缓解投资不足的问题;会计信息可比性与企业投资过度负相关,但相关关系不显著.文章从投资过度的影响因素以及样本规模的限制方面解释了不显著的原因,并指出后续研究可对影响投资过度的因素进行细分,扩大样本规模,做进一步的研究.     

基于神经网络的固体推进剂专家系统研究

固体推进剂研究对航天工业及武器研制具有重要的战略意义.鉴于其研制手段多以实验为主和分析软件智能化程度不高等因素,提出了基于神经网络的固体推进剂专家系统"SPES".神经网络可以较好的映射固体推进剂组成、结构及性能之间的复杂关系,因此SPES是基于材料设计思想和配方实例的高精度配方设计及性能预示软件.文中详细阐述了系统设计思想、总体结构、系统功能、关键技术及应用示例等方面.最后对SPES实现固体推进剂领域自动化和智能化给予了展望.     

金属多层的强度及界面强化能力研究进展

介绍了近年来内外有关金属多层强度及其界面强化能力研究的进展.结合近年来的研究结,分析并总结了金属多层强化的物理机制以及界面强化能力的,并对金属多层研究的发展趋势进行了展望.已有结表明,金属多层的强度通常随组元层的单层厚度减小而逐渐增大,并符合类似Hall-Petch的强化关系,但当单层厚度小于某一临界值以后,不同体系多层的强度与单层厚度之间偏离了Hall-Petch关系,表现出不同的变化趋势;金属多层的界面类及结构对其强化能力有显著影响,同时也影响金属多层的塑性变和断裂行为.     

具有连续分布时滞的双向联想记忆神经网络的全局稳定性

研究了具有连续分布式时滞的双向联想记忆神经网络的平衡点的全局稳定性问题. 在对神经元的激励函数的两种较宽松的假设下, 应用Lyapunov泛函法及不等式分析技巧获得了全局渐近稳定性的两个新的充分条件. 该结果推广了有关文献中已有的结论, 数值仿真的例子证明了其有效性.     

基于梁-颗粒模型的钢筋混凝土结构爆炸破坏数值研究

钢筋混凝土在爆炸载荷下的力学行为研究是防灾减灾与防护工程领域的重要研究课题.在离散元框架内,开发了模拟钢筋混凝土结构;在爆炸载荷下破坏过程的三维梁-颗粒模型,在弹性范围内,用矩阵位移法描述梁的变形与受力的关系,提出了用应力表达梁的强度准则.基于Cowper-Symonds理论,开发了描述钢筋在高加载率荷载下变形的梁-颗粒模型.采用C＋＋语言开发了模拟程序.进行了钢筋混凝土板在爆炸载荷下破坏规律的实验研究,同时用开发的模型进行了数值模拟,模拟结果与实验结果的对比表明：模拟结果与实验结果基本一致,所开发的模型可以反应爆炸载荷下钢筋混凝土材料的破坏特征,能够真实地体现爆坑的形成、裂纹的扩展、层裂等现象.     

A deep residual compensation extreme learning machine and applications

The extreme learning machine (ELM) is a type of machine learning algorithm for training a single hidden layer feedforward neural network. Randomly initializing the weight between the input layer and the hidden layer and the threshold of each hidden layer neuron, the weight matrix of the hidden layer can be calculated by the least squares method. The efficient learning ability in ELM makes it widely applicable in classification, regression, and more. However, owing to some unutilized information in the residual, there are relatively huge prediction errors involving ELM. In this paper, a deep residual compensation extreme learning machine model (DRC-ELM) of multilayer structures applied to regression is presented. The first layer is the basic ELM layer, which helps in obtaining an approximation of the objective function by learning the characteristics of the sample. The other layers are the residual compensation layers in which the learned residual is corrected layer by layer to the predicted value obtained in the previous layer by constructing a feature mapping between the input layer and the output of the upper layer. This model is applied to two practical problems: gold price forecasting and airfoil self-noise prediction. We used the DRC-ELM with 50, 100, and 200 residual compensation layers respectively for experiments, which show that DRC-ELM does better in generalization and robustness than classical ELM, improved ELM models such as GA-RELM and OS-ELM, and other traditional machine learning algorithms such as support vector machine (SVM) and back-propagation neural network (BPNN).     

基于RS GA ELM的矿井瓦斯涌出量预测

为了对煤矿井下瓦斯涌出量进行预测,采用粗糙集与改进极限学习机相结合的方法,在样本数据的筛选上吸取粗糙集数据约简的优点,充分利用极限学习机训练速度快、具有良好泛化性能的特点,并结合遗传算法选择最优的输入权值矩阵和隐含层偏差,避免随机产生所造成的误差。利用编写程序确定隐含层神经元个数,比依靠经验更为准确。在实际应用中选取煤层瓦斯含量、煤层埋藏深度、煤层厚度、煤层间距、工作面日产量五个因素作为预测的影响参数。研究结果表明:该预测模型预测的最大相对误差为5687 1%,最小相对误差为0,平均相对误差为2582 7%,相比改进前的预测模型具有更强的泛化能力和更高的预测精度。     

The size of motor units in laryngeal muscles of the rat

Summary The size of motor units in rat laryngeal muscles was determined by correlating the number of neurons labeled by i.m. injections of horseradish peroxidase with the number of motor end plates stained for acetylcholinesterase. The cricothyroid has a motor unit size of 8 muscle fibers per motor neuron and the posterior cricoarytenoid 4–5 muscle fibers per motor neuron.     

Neural network structure identification in inflation forecasting

Neural networks (NNs) are appropriate to use in time series analysis under conditions of unfulfilled assumptions, i.e., non‐normality and nonlinearity. The aim of this paper is to propose means of addressing identified shortcomings with the objective of identifying the NN structure for inflation forecasting. The research is based on a theoretical model that includes the characteristics of demand‐pull and cost‐push inflation; i.e., it uses the labor market, financial and external factors, and lagged inflation variables. It is conducted at the aggregate level of euro area countries from January 1999 to January 2017. Based on the estimated 90 feedforward NNs (FNNs) and 450 Jordan NNs (JNNs), which differ in variable parameters (number of iterations, learning rate, initial weight value intervals, number of hidden neurons, and weight value of the context unit), the mean square error (MSE), and the Akaike Information Criterion (AIC) are calculated for two periods: in‐the‐sample and out‐of‐sample. Ranking NNs simultaneously on both periods according to either MSE or AIC does not lead to the selection of the ‘best’ NN because the optimal NN in‐the‐sample, based on MSE and/or AIC criteria, often has high out‐of‐sample values of both indicators. To achieve the best compromise solution, i.e., to select an optimal NN, the preference ranking organization method for enrichment of evaluations (PROMETHEE) is used. Comparing the optimal FNN and JNN, i.e., FNN(4,5,1) and JNN(4,3,1), it is concluded that under approximately equal conditions, fewer hidden layer neurons are required in JNN than in FNN, confirming that JNN is parsimonious compared to FNN. Moreover, JNN has a better forecasting performance than FNN.     

Determination of the connectivity of newborn neurons in mammalian olfactory circuits

The mammalian olfactory bulb is a forebrain structure just one synapse downstream from the olfactory sensory neurons and performs the complex computations of sensory inputs. The formation of this sensory circuit is shaped through activity-dependent and cell-intrinsic mechanisms. Recent studies have revealed that cell-type specific connectivity and the organization of synapses in dendritic compartments are determined through cell-intrinsic programs already preset in progenitor cells. These progenitor programs give rise to subpopulations within a neuron type that have distinct synaptic organizations. The intrinsically determined formation of distinct synaptic organizations requires factors from contacting cells that match the cell-intrinsic programs. While certain genes control wiring within the newly generated neurons, other regulatory genes provide intercellular signals and are only expressed in neurons that will form contacts with the newly generated cells. Here, the olfactory system has provided a useful model circuit to reveal the factors regulating assembly of the highly structured connectivity in mammals.     

Signaling in the Chemosensory Systems

The small nematode Caenorhabditis elegans lives in the soil, where mechanical, thermal and most of all chemical stimuli strongly influence its behavior. Here we briefly review how chemical sensitivity is organized at the cellular and molecular level in this organism. C. elegans has less than 40 chemosensory neurons. With few exceptions each neuron senses more than one substance and each substance is sensed by more than one neuron. At the molecular level, as in other organisms, also in C. elegans, seven transmembrane G-protein-coupled receptors (GPCRs), heterotrimeric G proteins, cyclic nucleotidegated ion channels, TRP channels and Ca++ play crucial roles in chemical sensitivity. An unusual feature, possibly due to C. elegans's strong dependence on chemical cues, is the very large number of GPCR chemoreceptor genes (1300-1700) coded in its genome. Genetic approaches have also allowed the identification of new molecules involved in chemical sensitivity that would not have been discovered otherwise. In addition to the basic factors involved in primary signalling, the studies in C. elegans have revealed a network of regulatory pathways and molecules suggesting that fine modulation of the responsiveness of neurons is important, possibly to allow worms to negotiate a continuously changing environment. The experimental versatility of C. elegans has made it possible, in many cases, to determine precisely in which neuron a given molecule or pathway is required and for which biological response. This type of information can contribute to the general field of sensory signalling because it provides correlations between the biochemical properties of molecules and their cellular functions and between these and the in vivo behavioral responses of the animal.     

Principles of neural cell migration

Summary A basic property of immature neurons is their ability to change position from the place of their final mitotic division in proliferative centers of the developing brain to the specific positions they will occupy in a given structure of the adult nervous system. Proper acquisition of neuron position, attained through the process of active migration, ultimately affects a cell's morphology, synaptic connectivity and function. Although various classes of neurons may use different molecular cues to guide their migration to distant structures, a surface-mediated interaction between neighboring cells is considered essential for all types of migration. Disturbance of this cell-cell interaction may be important in several congenital and/or acquired brain abnormalities. The present article considers the basic mechanisms and principles of neuronal cell migration in the mammalian central nervous system.     

Building neural network models for time series: a statistical approach

This paper is concerned with modelling time series by single hidden layer feedforward neural network models. A coherent modelling strategy based on statistical inference is presented. Variable selection is carried out using simple existing techniques. The problem of selecting the number of hidden units is solved by sequentially applying Lagrange multiplier type tests, with the aim of avoiding the estimation of unidentified models. Misspecification tests are derived for evaluating an estimated neural network model. All the tests are entirely based on auxiliary regressions and are easily implemented. A small‐sample simulation experiment is carried out to show how the proposed modelling strategy works and how the misspecification tests behave in small samples. Two applications to real time series, one univariate and the other multivariate, are considered as well. Sets of one‐step‐ahead forecasts are constructed and forecast accuracy is compared with that of other nonlinear models applied to the same series. Copyright © 2006 John Wiley & Sons, Ltd.     

Optical monitoring of activity of many neurons in invertebrate ganglia during behaviors

Optical methods for monitoring neuron activity were developed because these methods lend themselves to simultaneous multiple-site measurements. With the use of new voltage-sensitive dyes, the dye-related pharmacology and photodynamic damage appear to be relatively unimportant. Using multiple-site measurements made with a 124-element photodiode array, we estimated that approximately 30 of the 200 neurons present in the Navanax buccal ganglion make action potentials during feeding and that approximately 300 of the 1100 neurons present in the Aplysia abdominal ganglion are active during the gill-withdrawal reflex. The fact that a light mechanical touch to the siphon skin activated such a large number of neurons in the abdominal ganglion suggests that understanding the neuronal basis of the gill-withdrawal reflex and its behavioral plasticity may be forbiddingly difficult.     

Optical monitoring of activity of many neurons in invertebrate ganglia during behaviors

Summary Optical methods for monitoring neuron activity were developed because these methods lend themselves to simultaneous multiple-site measurements. With the use of new voltage-sensitive dyes, the dye-related pharmacology and photodynamic damage appear to be relatively unimportant. Using multiple-site measurements made with a 124-element photodiode array, we estimated that approximately 30 of the 200 neurons present in theNavanax buccal ganglion make action potentials during feeding and that approximately 300 of the 1100 neurons present in theNavanax buccal ganglion make are active during the gill-withdrawal reflex. The fact that a light mechanical touch to the siphon skin activated such a large number of neurons in the abdominal ganglion suggests that understanding the neuronal basis of the gill-withdrawal reflex and its behavioral plasticity may be forbiddingly difficult.