Java语言与人工神经网络应用

从控制工程的角度,概括介绍了作为国际互联网的最新核心技术之一的Ｊａｖａ语言的特点和发展,并具体结合人工神经网络的仿真与实现,以实际例子论述其在科学研究中将发挥的巨大效用和带来的新思想．     

Marked electrical synapses on AP neuron of the leech (Whitmania pigra)

An initial attempt reveals parameters of marked electric synapse on AP perceptive neuron of leech. The width of gap between marked AP and either neuro-or glio-process is 2–3 nm, the diameter of the hexagonal array of connexion is 8–9 nm with a central pole of 2–2.5 nm, the same as the structure of annular lamellar bodies internalized by AP-glia processes in the AP.     

利用神经网络预测混凝土的抗压强度

应用人工神经中的神经元网络理论,预测混凝土的抗压强度值,探讨出一条预测混凝土抗压强度的理论方法,其结果具有较高的精确度。     

自适应神经网络PID控制器

提出了一种新型神经网络PID控制器，其学习速率是通过三层前向BP网络在线辨识学习，使神经元有较强的智能性、自适应和自学习的能力；同时，将Smith预估器与神经元PID控制器相结合，能更有效地抑制纯滞后的影响。仿真结果表明该控制器有较好的控制效果和鲁棒性。     

微小型LTA飞行机器人主动避障控制研究

采用基于行为的研究方法，设计了一种基于电机神经元网络的微小型LTA飞行机器人主动避障控制方法。该方法通过电机神经元网络选取飞行机器人的飞行动作，采用增强式学习方法对飞行结果作出评判，利用评判结果在线修改网络权值，从而实现微小型LTA飞行机器人在未知、动态环境中的主动避障。飞行机器人的仿真运行分析，验证了该方法的有效性。     

基于模拟电路的无刷直流电动机神经元控制器

从神经元特性出发，研究一种适用于无刷直流电动机控制的模拟电路神经元控制器，获得连续的神经元控制器表达式和权值变化表达式，并进行原理框图设计和给出具体的权值学习电路，仿真实验表明，该控制器具有良好的控制品质，证明了该系统的有效性。     

一种基于专家系统的变结构神经元自适应PID控制器的设计与实现

根据温度控制系统的具体特性提出一种变结构神经元自适应PID控制器算法,特别是在调整神经元控制器最敏感的参数K值时引入了专家经验.实验结果表明,利用这种控制器进行电加热炉温度控制具有调节时间短、稳定性能好、遇干扰回复能力强等优点.     

Differential effects of the serotonin receptors on cultured rat cerebral cortical neurons

Effects of serotonin (5-HT) on cerebral cortical neurons were examined by patch clamp techniques. 5-HT produced a variety of responses such as outward (19/73 patches/neurons), slow inward (15/73 patches/neurons), fast inward (8/73 patches/neurons), and mixed currents (initially fast inward deflection followed by an outward response: 2/73 patches/neurons), with a latency of 12 sec, 15 sec, 0 sec, and 0 sec respectively, at a holding potential of −60 mV in whole-cell patches. The fast inward currents were again evoked by a selective 5-HT3 receptor agonist, 1-(m-chlorophenyl)-biguanide hydrochloride (CPBG). In the cell-attached patch clamp configuration, 5-HT inside the patch pipette elicited single channel currents with slope conductances of 42 pS and 132 pS (4/42 patches/neurons). CPBG inside the patch pipette evoked inward single channel currents with a lower slope conductance of 41 pS (3/23 patches/neurons). In contrast, application of 5-HT or a 5-HT₂ receptor agonist, α-methyl-5-hydroxytryptamine-maleate, outside the patch pipette induced outward single channel currents with a major slope conductance of 140 pS (8/30 patches/neurons) or 135 pS (6/20 patches/neurons), respectively. These results indicate that the outward and fast inward currents may be mediated respectively by the 5-HT₂ receptor, which is coupled to a G-protein, and by the 5-HT₃ receptor, which contains the non-selective cation channel, and that the mixed type may be caused by both the 5-HT₂ and 5-HT₃ receptors. Received 27 September 1996; received after revision 4 November 1996; accepted 7 November 1996     

鼠耳蝠下丘听神经元对超声刺激的反应特性

实验在５只鼠耳蜗上进行，共观察了２２７个对超声刺激发生反应的下丘听神经元。这些神经元听反应的最佳频率在１４．３－７６．２ｋＨｚ之间；潜伏期在３．０－１２．０毫秒之间，多数神经元（７９．３％）为５．０－７．０毫秒；最低阈值位于－２０．０－７０．０ｄＢ ＳＰＬ之间；多数神经元（８２．０５％）的调谐曲线为宽阔型，少数（１７．９５％）为狭窄型；反应最佳频率沿下丘背腹轴呈明显有序地排列。     

一种单神经元自抗扰控制器

针对自抗扰控制器可调参数多且不易整定的问题,提出了一种用单神经元改进非线性状态误差反馈控制律的算法.利用神经网络的自学习能力,采用一个单神经元构造自适应参数,使参数依据系统误差的变化自动作相应地调整,从而完成参数的在线自整定.仿真结果表明,改进后的控制器调整参数大大减少,而且具有更强的适应性和鲁棒性.