Pattern discrimination and firing behavior of cold receptor model aided with noise

A method for characterizing and identifying firing patterns of neural spike trains is presented. Three characteristic variables defined at sequential moments, including two formal derivatives and the integration of the counting process, are introduced to reflect the temporal patterns of a spike train. This paper also examines how noise interacts with encoding mechanisms of neuronal stimulus in a cold receptor. From ISI series and bifurcation diagrams it is shown that there are considerable differences in interval distributions and impulse patterns caused by purely deterministic simulations and noisy simulations. The ISI-distance can be used as an effective and powerful way to measure the noise effects on spike trains quantitatively. It is found that spike trains observed in cold receptors can be more strongly affected by noise for low temperatures than for high temperatures in some aspects; meanwhile, spike train has greater variability with the noise intensity increasing.     

方钢管混凝土短柱轴心受压承载力的神经网络模拟

由于方钢管混凝土的侧向约束机构复杂,对方钢管混凝土柱强度承载力的计算至今仍没有一种统一的方法。本文拟采用神经网络方法对轴心受压方钢管混凝土短柱的承载力进行模拟。以混凝土抗压强度、钢管的屈服强度、套箍指标、截面尺寸和宽厚比等五个参数为网络输入,以构件的极限承载力为网络输出,构建多层前馈神经网络来描述它们之间的非线性关系。利用55组试验数据对网络进行训练和测试,并将其预测值与三种承载力计算模型的预测值进行比较。对比结果表明本文建立的神经网络模型对55组试验数据给出了最好的模拟精度,可作为预测方钢管混凝土柱承载能力的一种新方法。     

基于顺序雨流法和临界平面法的多轴变幅低周疲劳寿命评价

 在多轴变幅疲劳寿命预测过程中,合适的雨流计数法对复杂加载历程分析非常重要,但是大多数雨流计数过程往往无法保持原始的加载顺序特性,进而会导致非保守的疲劳损伤和寿命预测．本文提出一种考虑加载顺序效应并基于临界面概念的多轴实时顺序雨流法,该方法既具有实时顺序计数特点,同时与Bannantine-Socie多轴雨流法结合,可以实现对主要通道内的载荷历程实时的雨流计数．基于Morrow 模型,提出一种新的考虑加载顺序的线性损伤累积方法．相对于传统雨流计数法需要得到完整的载荷数据后才能进行分析的特点,新方法计算效率更高,实用性更强．通过对316L 不锈钢的多轴疲劳试验数据的分析,验证了该方法在多轴疲劳寿命预测过程中的有效性．     

Classifying wakes produced by self-propelled fish-like swimmers using neural networks

We consider the classification of wake structures produced by self-propelled fish-like swimmers based on local measurements of flow variables. This problem is inspired by the extraordinary capability of animal swimmers in perceiving their hydrodynamic environments under dark condition. We train different neural networks to classify wake structures by using the streamwise velocity component, the crosswise velocity component, the vorticity and the combination of three flow variables, respectively. It is found that the neural networks trained using the two velocity components perform well in identifying the wake types, whereas the neural network trained using the vorticity suffers from a high rate of misclassification. When the neural network is trained using the combination of all three flow variables, a remarkably high accuracy in wake classification can be achieved. The results of this study can be helpful to the design of flow sensory systems in robotic underwater vehicles.     

脉动型神经元网络的联想记忆与分割

以广泛讨论的Hodgk in-Hux ley神经元节点组成脉动神经元网络,从神经系统空时模式编码理论研究网络的记忆(或模式)存储与时间分割问题。给定一个输入模式,它是几种模式的叠加,网络能够以一部分神经元同步发放的形式一个接一个地在时间域分割出每一种模式。如果输入的模式是缺损的,系统能够把它们恢复成完好的原型,即神经网络的联想记忆功能。     

基于Elman型神经网络的金川二矿地表岩移时序预测 模型

建立于煤矿开采基础之上的矿山开采沉陷理论和预测方法并不适用于象金川这样厚大、陡倾的金属矿床开采的岩移问题,因此,本文探讨利用神经网络来对地表岩移进行预测。根据Elman神经网络能够逼近任意非线性函数的特点和具有反映系统动态特性的能力,提出了利用Elman神经网络建立地表岩移时序预报模型的方法。利用金川二矿区GPS监测所得到的时间序列数据,通过对Elman神经网络模型预测值与GPS实测值之间的比较,结果表明模型预测显示了良好的准确性,特别是在时间步长较短情况下,应用于实际预测一定程度上可以弥补金属矿山岩移预测方法不足的缺憾。     

Modeling and prediction of chaotic systems with artificial neural networks

This study of chaotic systems and their prediction is motivated by the fact that many phenomena, both natural and man‐made, are of a chaotic nature. Such phenomena include but are not limited to earthquakes, laser systems, epileptic seizures, combustion, and weather patterns. These phenomena have previously been thought to be unpredictable. However, it is indeed possible to predict time series generated by chaotic systems. The primary objective of this study is to develop a system that would train the artificial neural network (ANN) and then predict the future data of the process. In the present application, the chosen chaotic data set was obtained by solving Lorenz's equations. To predict the future data, the concept of a multilayer feed‐forward ANN with nonlinear auto‐regressive moving averages with exogenous input is used. A Backpropagation algorithm is used to train the network for the chaotic data. The final updated weights from the trained network were then used for the prediction of the future values of the system. Lyapunov exponents, phase diagrams and statistical analyses were used to evaluate the neural network output. A correlation of 94% and a negative Lyapunov exponent indicate that the results obtained from ANN are in good agreement with the actual values. Copyright © 2009 John Wiley & Sons, Ltd.     

A fully discrete ALE method over untwisted time–space control volumes

In this paper, a fully discrete high‐resolution arbitrary Lagrangian–Eulerian (ALE) method is developed over untwisted time–space control volumes. In the framework of the finite volume method, 2D Euler equations are discretized over untwisted moving control volumes, and the resulting numerical flux is computed using the generalized Riemann problem solver. Then, the fluid flows between meshes at two successive time steps can be updated without a remapping process in the classic ALE method. This remapping‐free ALE method directly couples the mesh motion into a physical variable update to reflect the temporal evolution in the whole process. An untwisted moving mesh is generated in terms of the vorticity‐free part of the fluid velocity according to the Helmholtz theorem. Some typical numerical tests show the competitive performance of the current method. Copyright © 2016 John Wiley & Sons, Ltd.     

Understanding neuromotor strategy during functional upper extremity tasks using symbolic dynamics

The ability to model and quantify brain activation patterns that pertain to natural neuromotor strategy of the upper extremities during functional task performance is critical to the development of therapeutic interventions such as neuroprosthetic devices. The mechanisms of information flow, activation sequence and patterns, and the interaction between anatomical regions of the brain that are specific to movement planning, intention and execution of voluntary upper extremity motor tasks were investigated here. This paper presents a novel method using symbolic dynamics (orbital decomposition) and nonlinear dynamic tools of entropy, self-organization and chaos to describe the underlying structure of activation shifts in regions of the brain that are involved with the cognitive aspects of functional upper extremity task performance. Several questions were addressed: (a) How is it possible to distinguish deterministic or causal patterns of activity in brain fMRI from those that are really random or non-contributory to the neuromotor control process? (b) Can the complexity of activation patterns over time be quantified? (c) What are the optimal ways of organizing fMRI data to preserve patterns of activation, activation levels, and extract meaningful temporal patterns as they evolve over time? Analysis was performed using data from a custom developed time resolved fMRI paradigm involving human subjects (N=18) who performed functional upper extremity motor tasks with varying time delays between the onset of intention and onset of actual movements. The results indicate that there is structure in the data that can be quantified through entropy and dimensional complexity metrics and statistical inference, and furthermore, orbital decomposition is sensitive in capturing the transition of states that correlate with the cognitive aspects of functional task performance.     

基于PTCNN的结构布局优化问题研究

对结构优化中的难点布局优化问题,基于交叉学科的优势,运用脉冲暂态混沌神经网络（PTCNN）的方法,对一个新型布局优化模型进行了寻优计算。在PTCNN算法中应用脉冲混沌动力自然解决了杆件的恢复和删除问题;利用神经网络的多神经元并行构成特点,通过参数调整方法解决了不同数量级变量问的耦合问题。因此PTCNN不仅是模型的寻优算法,更成为布局优化的一部分。此外构造了一个应力关联系数σ_AI,使面积变量随杆中应力大小按比例自适应下降。算例结果表明,基于PTCNN方法解决结构布局优化问题,有效且具有自适应性,布局优化效果明显。     

大脑血液动力学现象中的能量编码

神经信息的编码与解码是神经科学中的核心研究内容,同时又极具挑战性.传统的编码理论都具有各自的局限性,很难从脑的全局运行方式上给出有效的理论.而由于能量是一个标量具有可叠加性,因此能量编码理论可以从神经元活动的能量特征出发来研究脑功能的全局神经编码问题,取得了一系列的研究成果.本研究以王-张神经元能量计算模型为基础,构建了一个多层次结构的神经网络,通过计算机数值模拟得到了神经网络的能量消耗和血液中葡萄糖供能的变化情况.计算结果显示,和网络的神经活动达到峰值的时间相比,血液中葡萄糖的供能达到峰值的时间延迟了约5.6s.从定量的角度再现了功能性核磁共振(fMRI)中的血液动力学现象:大脑某个脑区的神经元集群被激活以后经过5~7 s的延迟,脑血流的变化才会大幅增加.模拟结果表明先前发表的由王-张神经元模型所揭示的负能量机制在控制大脑的血液动力学现象中起着核心的作用,预测了刺激条件下大脑的能量代谢与血流之间变化的本质是由神经元在发放动作电位过程中正、负能量之间的非平衡、不匹配性质所决定的.本文的研究结果为今后进一步探究血液动力学现象的生理学机制提供了新的研究方向,在神经网络的建模与计算方面给出了一个新的视角和研究方法.      

采用两步优化器的深度配点法与深度能量法求解薄板弯曲问题

随着计算机技术的进步以及机器学习算法的进一步发展,深度学习方法逐渐被广泛引用于各行各业中。本文发展并比较了适应于工程计算的深度配点法与深度能量法并应用于求解薄板弯曲问题。深度配点法采用物理驱动的深度神经网络来,并将物理信息（偏微分方程强形式）引入到损失函数中,最终将求解薄板弯曲问题简化为优化问题。深度能量法则是采用系统总势能驱动的神经网络。根据最小势能原理,在所有的可能位移场中,真实位移场的总势能取最小值,因此我们可以使用总势能构造损失函数,从而求解薄板弯曲问题。对于边界条件,通过罚函数法将有约束最优化问题转化为求解无约束最优化问题。深度配点法与深度能量法的适用性基于神经网络的通用近似定理。由于物理信息跟总势能的引入,增加了神经网络训练的困难,为了解决这个问题,我们发展了两步优化器方法。数值结果表明,深度配点法与深度能量法很适合求解薄板弯曲问题,并且程序实现简单,实现了真正意义上的“无网格法”。     

神经网络状态方程在强爆炸冲击波数值模拟中的应用

强爆炸数值模拟的主要挑战在于如何准确地描述爆炸产物状态方程。利用BP神经网络和强爆炸产物状态数据对神经网络产物状态方程进行训练,并将得到的状态方程植入自编的一维球对称数值模拟程序,对强爆炸冲击波参数进行了计算。结果显示,计算得到的冲击波峰值超压、冲击波到时、正压时间与标准值吻合较好,证明将神经网络状态方程应用于强爆炸冲击波数值模拟是可行的。研究结果对确定强爆炸数值模拟方法具有很好的借鉴意义。     

基于Elman神经网络和Kalman滤波的捷联惯导精对准

在采用Kalman滤波进行捷联惯导精对准时,当模型存在误差或系统噪声不能反映实际噪声时,会降低滤波精度甚至导致滤波发散.针对这个问题,提出基于Elman神经网络和Kalman滤波的捷联惯导精对准方法.首先对已知噪声统计特性的系统进行Kalman滤波,将稳定可靠的状态估值作为网络期望输出用来训练Elman网络,然后再用训练好的网络对未知噪声统计特性系统进行状态估计.利用仿真数据对该算法进行验证,结果表明该算法能够克服Kalman滤波精对准的缺陷,提高了对准精度,尤其是航向角的精度.     

Controllable spiking dynamics in cascaded VCSEL-SA photonic neurons

We propose a photonic neural system composed of three cascaded vertical-cavity surface-emitting lasers with an embedded saturable absorbers (VCSEL-SAs) and numerically investigate the encoding, propagation and storage characteristics of the spiking patterns in this system. The results show that, with suitable perturbation strength, the first VCSEL-SA (VCSEL-SA1) can convert the stimulus into spike response. Increasing both the perturbation strength and the bias current of active region is beneficial to improve the conversion rate. Moreover, the spiking patterns generated by VCSEL-SA1 can be stably propagated into another two VCSEL-SAs (VCSEL-SA2 and VCSEL-SA3) with a certain delay through adjusting the coupling weight. Additionally, after introducing a feedback into VCSEL-SA1, the fired spiking patterns can be successfully stored in this proposed system. The obtained results can offer great potential for future, brain-inspired ultrafast neuromorphic computing system.     

利用子波分析对平壁湍流猝发现象的研究

利用槽道湍流直接数值模拟的数据库,采用子波分析的方法。对平壁湍流猝发现象的多尺度特性进行了研究,在不同惊讶上对猝发平均周期进行了统计,并利用局部标度指数研究了猝发过程的奇异性。     

铁路货车纵向载荷的试验测量和统计推断研究

通过线路测试得到我国通用线路C70货车车钩纵向载荷时间历程,利用雨流计数法得到车钩纵向载荷均值-频次谱。分别对我国通用线路C70货车车钩纵向载荷谱、大秦线C80B货车车钩纵向载荷谱、AAR90.7t漏斗车车钩纵向载荷谱及FMG矿石车车钩纵向载荷谱的分布特征进行分析,并对最大纵向载荷进行统计推断研究。研究结果表明:铁路货车最大纵向载荷随着轴重、牵引吨位的增大而增大。经纵向动力学仿真分析发现,单车轴重、牵引吨位、编组方式的不同是导致上述变化的原因。     

遗传-神经网络算法优化飞机垂尾

通过优化垂尾复合材料蒙皮铺层体系,可以减少垂直尾翼的气动变形,进而改善飞机的方向安定性。采用遗传算法能够较好地分析这类复杂优化问题,但遗传算法的计算工作量大,计算所需时间过长。本文提出遗传一神经网络的算法,该算法首先采用神经网络预测遗传变量的值,进而缩短遗传变量的范围,然后采用遗传算法优化得到最优解。结果表明,这种方法在保证计算精度的情况下,减少了计算时间,提高了工作效率。     

等p路径十砂土本构关系的归一化特性及数值建模方法

用神经网络对土体进行建模能反映应力路径相关性、反映土的剪胀剪缩以及反映体应力、剪应力对体应变、剪应变的交互影响,因而成为一种比较理想的建模方式.能否在样本有限的情况下获得精度比较高的本构模型正是主要的研究目的.通过研究中密砂在等p路径下的三轴试验曲线,发现其应力-应变关系曲线在常规应力范围内具有归一化特性.选择合适的归一化指标对砂土三轴试验数据进行归一化,以归一化的试验数据为训练样本进行神经网络训练,得到了比较理想的砂土的神经网络本构模型.本构模型仿真值与试验值符合较好,表明所给出的建模方法是合理的.提出的建模方法可以在所有试验数据的基础上自动实现概率寻优,能有效降低噪声信号的干扰、减小试验数据的分散造成的影响.     

阵发型神经元网络的联想记忆与分割

以Hindmarsh-Rose(HR)神经元组成网络,研究这些网络的记忆(或模式)存储与时间分割问题．给定一个输入模式,它是几种模式的叠加,网络能够以一部分神经元同步发放的形式一个接一个地分割出每一种模式．如果输入的模式是缺损的,系统能够恢复为原型,即网络具有联想记忆功能．模拟也得到据作者所知至今还未报道的一些现象．