米帕明对大鼠皮层神经细胞缺血性损伤的保护作用

目的: 研究米帕明(Imi) 对培养大鼠神经细胞缺血和谷氨酸(Glu) 损伤的保护作用。方法: :分离培养 15～ 18 d 胎龄的大鼠神经细胞, 用连二亚硫酸钠消除培养基中的氧, 合并培养基缺糖模拟细胞缺血性损伤;加入Glu 模拟兴奋毒性损伤, 测定乳酸脱氢酶、一氧化氮、丙二醛和超氧化物歧化酶的含量变化, 观察缺血和Glu 对神经细胞的损伤及 Imi 的保护作用。结果: :缺血及 Glu 引起神经细胞明显损伤性变化, 死亡率升高, 培养上清液中乳酸脱氢酶、一氧化氮含量升高, 细胞匀浆中丙二醛生成增加, 超氧化物歧化酶含量明显减少。Imi 10^-8 ～ 10^-6 mol·L^-1能不同程度地减轻上述损伤性变化。结论: :Imi 对神经细胞缺血性损伤有保护作用, 机制可能与抗脂质过氧化及钙拮抗作用有关。     

环境评价的人工神经网络方法

简要讨论了神经网络的基本原理, 介绍了多层前馈神经网络模型及BP 算法解决问题的基本思路与方法, 分别对两个地方的大气质量及水质进行了评价, 同时分别采用模糊数学法及灰色聚类法对这两组数据进行了处理, 所得结果一致。     

FINITE ELEMENT SOLUTIONS WITH FEEDBACK NETWORK MECHANISM THROUGH DIRECT MINIMIZATION OF ENERGY FUNCTIONALS

In this paper, a new algorithm for solving a minimization problem of the functional of a variational principle was proposed by focusing on the state transition process of the interconnected neural network. A new concept of neuron called ‘feedback neuron’ was introduced into the conventional interconnected neural network. This method was readily applied to the finite element analyses of incompressible viscous flows. Furthermore, favourable feature of the neural network for parallel platform was explored on a massively parallel computer AP1000.     

风力机变桨系统单神经元自适应PID控制

在分析风力发电系统能量转换理论的基础上,将单神经元自适应算法与常规PID算法结合,通过单神经元自学习,对PID控制器的系数进行在线调整,提出一种单神经元自适应PID调节变桨角度的控制方法,建立了风力机桨距角控制数学模型.风速超过额定值后,利用单神经元自适应PID控制器进行变桨控制,将控制结果与采用常规PID控制器的控制性能进行了对比.结果表明:单神经元自适应PID控制器可以将输出功率恒定在额定值,能够对变桨系统进行有效控制,达到预期的目标.     

左乙拉西坦对急性癫痫小鼠海马GABA能神经元电生理的研究

目的: 研究癫痫小鼠海马CA1区γ-氨基丁酸(GABA)能神经元电生理变化及使用左乙拉西坦(levetiracetam,Lev)干预治疗后的变化。方法: 将FVB-Tg小鼠随机分为正常组、模型组、左乙拉西坦干预组。左乙拉西坦采用灌胃给药。在造模成功后直接断头取脑,制作脑片,在人工脑脊液中孵育后,利用膜片钳技术,进行全细胞记录。结果: 与正常组相比,模型组动作电位的绝对不应期延长(P<0.05),电位能障(Vts-Vr)值上升(P<0.05),动作电位间距延长(P<0.05),差异有统计学意义。与模型组相比,左乙拉西坦干预组动作电位的绝对不应期缩短(P<0.01),电位能障(Vts-Vr)值下降(P<0.01),动作电位间距缩短(P<0.01),差异有显著统计学意义。结论: 左乙拉西坦组干预后癫痫小鼠海马CA1区GABA能神经元内在特性和编码能力增强,GABA能神经元功能增加可能在癫痫的治疗中起重要作用。     

循环神经网络前馈补偿的压电驱动器跟踪控制

压电驱动器固有的迟滞特性,以及其他动态特性严重地影响其跟踪性能.循环神经网络能够准确拟合非线性系统,并且具有记忆存储能力,本文设计了一种循环神经网络对压电驱动器的迟滞特性进行建模,进而得到能够准确模拟输出位移和输入电压之间关系的逆模型,并据此对压电驱动器进行前馈补偿.此外,考虑到建模误差以及其他扰动对驱动器跟踪精度的影...     

基于低时延和高精度脉冲神经网络的目标检测

人工神经网络(artificial neural network, ANN)在众多领域取得了显著进展, 但其对计算资源和能耗的高需求限制了其在硬件端的部署和应用. 脉冲神经网络(spiking neural network, SNN)因其低功耗和快速推理的特性, 在神经形态硬件上表现出色. 然而, SNN的神经元动态和脉冲发放机制导致其训练过程复杂, 目前主要研究集中在图像分类任务上, 本文尝试将SNN应用于更为复杂的计算机视觉任务. 本文以YOLOv3-tiny网络为基础, 提出了Spiking YOLOv3模型, 其符合SNN特性的网络模型, 在检测任务上实现了更高的准确度, 并将平均推理时间减少至约原来工作的1/4. 此外, 我们还分析了ANN-SNN转换过程中产生的转换误差, 并采用量化激活函数对Spiking YOLOv3模型进行了优化以减小转换误差. 优化后的模型平均推理时间减少至约原来的1/2, 并在VOC与UAV数据集上实现在ANN-SNN无损转换, 显著提升了基于该模型的检测效率.     

Effects of channel blocks on the spiking regularity in clustered neuronal networks

In this paper,we discuss the influences of channel blocks on the spiking regularity in a clustered neuronal network by applying stochastic Hodgkin-Huxley neuronal models as the building blocks.With the aid of simulation results,we reveal that the spiking regularity of the clustered neuronal network could be resonantly enhanced via fine-tuning of the non-blocked potassium channel fraction xK.While the non-blocked sodium channel fraction xNa can enhance the spiking regularity of the clustered neuronal network in most cases.These results indicate that not only sodium channel blocks but also potassium channel blocks could have great influences on the regularity of spike timings in the clustered neuronal networks.Considering the importance of spike timings in neuronal information transforming processes,our results may give some implications for understanding the nonnegligible role of randomness in ion channels in neuronal systems.     

The bridge relating process neural networks and traditional neural networks

The process neural network (PrNN) is an ANN model suited for solving the learning problems with signal inputs, whose elementary unit is the process neuron (PN), an emerging neuron model. There is an essential difference between the process neuron and traditional neurons, but there also exists a relation between them. The former can be approximated by the latter within any precision. First, the PN model and some PrNNs are introduced in brief. And then, two PN approximating theorems are presented and proved in detail. Each theorem gives an approximating model to the PN model, i.e., the time-domain feature expansion model and the orthogonal decomposition feature expansion model. Some corollaries are given for the PrNNs based on these two theorems. Thereafter, simulation studies are performed on some simulated signal sets and a real dataset. The results show that the PrNN can effectively suppress noises polluting the signals and generalize quite well. Finally some problems on PrNNs are discussed and further research directions are suggested.