酸碱中和反应中pH值的单神经元自适应控制研究

基于酸碱中和反应过程中pH值变化的动态模型,设计单神经元自适应控制器,解决化工生产过程中pH值的PID控制所存在的调节时间长、超调量大和系统抗干扰性差的问题。在相同条件下与PID控制器进行对比实验,实验结果表明单神经元自适应控制器明显改善pH值过程控制系统的动态性能,该控制系统具有较强的自适应性与鲁棒性。     

电液位置伺服系统的单神经元PSD控制研究

研究了单神经元PSD控制器的一般结构，探讨了自适应PSD算法和改进的神经元学习规则，在SIMULINK环境下，针对实际的电液位置伺服系统进行了动态仿真，结果表明单神经元PSD控制器能使系统得到较好的控制效果。     

基于队列长度和速率的拥塞控制神经网络方法

文中研究了网络拥塞控制问题。PID控制器是实现网络拥塞控制非常有效的方法,能够实现对网络的主动队列管理。文中根据队列长度和变化速率,利用神经网络实现传统的比例微分积分器（PID）功能,从而提出了基于队列长度和速率的拥塞控制神经网络方法（RSPID）。该方法利用神经网络的加权动量梯度学习算法,自动调节控制参数,克服了传统PID控制方法由于控制器参数固定带来的适应性和稳定性问题。仿真结果表明,RSPID算法的鲁棒性和队列长度性能要优于PID算法。     

Enantiomeric Composition of Monoterpene Hydrocarbons in Some Conifers and Receptor Neuron Discrimination of α-Pinene and Limonene Enantiomers in the Pine Weevil, Hylobius abietis

The enantiomeric composition of seven monoterpene hydrocarbons in headspace volatiles of spruce sawdust and seedlings (Picea abies), pine seedlings (Pinus sylvestris), and branches of juniper (Juniperus communis) was determined by gas chromatographic separation on a -Cyclodextrin column. For the six monoterpenes, -pinene, camphene, -pinene, sabinene, limonene, and -phellandrene, both enantiomers were present, whereas for 3-carene only the (+)-configuration was found. The amount of each enantiomer varied considerably both in relation to total amount of all of them, and for the six pairs also in relation to the opposite enantiomer. One olfactory receptor neuron in the pine weevil (Hylobius abietis) showed a strong response to -pinene when stimulated with all four headspace materials via a GC equipped with a DB-WAX column. The same neuron was subsequently tested with repeated stimulations via the GC effluent containing the (+)- or (–)-enantiomer. A marked better response to (+)- than to (–)--pinene was elicited. Another olfactory receptor neuron that responded strongly to limonene when stimulated with the spruce volatiles was tested for enantiomers of limonene. This neuron responded more strongly to (–)- than to (+)-limonene, when stimulated alternately with each of the limonene enantiomers. Discrimination between enantiomers by plant olfactory receptor neurons suggests that the enantiomeric ratios of volatile compounds may be important in host location by the pine weevil.     

用于队列管理的模糊增益神经元自适应控制器

针对具有参数时变及非线性特性的网络拥塞控制系统,提出了一种模糊增益神经元主动队列管理算法(FNAQM). 采用路由器队列长度及数据流速作为拥塞度量,在检测当前拥塞信息的同时,预测未来拥塞的状况. 结合神经元控制和模糊控制的优点,利用单神经元计算数据包标记概率,采用有监督的Hebb学习规则在线调整加权系数. 设计的模糊控制器可动态调整神经元增益,能获得更好的控制性能. FNAQM具有结构简单、易于实现、自适应能力强等优点. 仿真实验结果表明,FNAQM能快速将队列调整至目标值,并维持较小的队列抖动,对动态数据流和非响应流具有良好的鲁棒性.     

感知神经网络的信息传输及电路实现研究

为提高神经元网络信号的传输效率,改善人们对外界刺激的感知能力,本文采用平均互信息量的评价方法,研究了费茨休夫-南云感知神经元模型中信息传输的随机共振现象,探讨了并联单层神经元网络和双层神经元网络的信息处理功能,并用PSpice电路对神经元网络的信息传输进行了实验验证。数值模拟与电路仿真结果表明,FHN神经元并联模型和双层神经元网络具有随机共振现象,在一定的噪声强度范围内,双层神经元网络对输入信号的响应好于单层神经元网络,噪声能够提高神经元网络的输入输出互信息量,平均互信息量的最大值对应最优的传输性能,证明噪声对神经网络信息传输具有增强作用。该研究对感知神经元群体的信息传输具有重要的启示意义。     

基于PCNN的自动图像分割

脉冲耦合神经网络(PCNN)模型在数字图像分割中得到了广泛的应用,但对网络参数的确定以及最优结果的选则一直是一个难题,主要是以人工经验为主,虽然提出了一些参数自动设置的算法,但都缺乏对模型本身数学理论的研究和分析.通过对PCNN网络神经元点火特性的分析,提出了一种基于图像最大灰度值最小时刻点火的参数自适应设定算法,并针对该算法构造了一种新的用于最优结果选择的判定准则.将其用于lena等图像分割中,取得了与主观评价相一致的结果,而且表现出更快的速度和较好的鲁棒性.     

Probabilistic Bit Stream Neural Chip: Theory

A design for a neural network chip is proposed using probabilistic bit streams to represent real values. This paper analyzes the performance of the proposed neurons in this design and demonstrates that very simple operations can be used to obtain the desired functionality. It is also shown that a suitable ‘activation function’ for neurons of this type can be obtained using the interaction of two probability distributions. Finally, the paper introduces a variant of the back-propagation learning algorithm which involves computing the derivatives of the output with respect to individual weights in a network of such units.     

基于模糊单神经元PID的比例同步控制研究北大核心

针对传统PID算法在液压同步系统中整定困难、稳定性不佳、同步效果不理想的问题,提出一种基于模糊单神经元PID复合算法的双缸耦合同步控制策略,提高同步系统的性能。建立比例阀控非对称缸双缸同步系统数学模型;利用Simulink软件,对比分析传统PID算法、单神经元PID算法和模糊单神经元PID算法的控制效果;最后通过试验验证3种控制策略下的系统同步性能。试验结果表明:使用模糊单神经元PID的耦合控制策略,系统同步误差明显减小,响应速度更快,设计的控制策略有效。     

Stability Analysis for Delayed Neural Networks with Discontinuous Neuron Activations

This paper considers a class of delayed neural networks with discontinuous neuron activations. Based on the theory of differential equations with discontinuous right‐hand sides, some novel sufficient conditions are derived that ensure the existence and global exponential stability of the equilibrium point. Moreover, by adopting the concept of convergence in measure, convergence behavior for the output is discussed. The obtained results are independent of the delay parameter and can be thought of as a generalization of previous results established for delayed neural networks with Lipschtz continuous neuron activations to the discontinuous case. Finally, we give a numerical example to illustrate the effectiveness and novelty of our results by comparing our results with those in the early literature.