A circuit technique for broadbanding the electronic tuning range of Gunn oscillators

A circuit technique is described whereby the electronic tuning range obtained by varactor tuning solid-state oscillators, such as Gunn oscillators, can be improved. The principle of the technique has been demonstrated by doubling the tuning range obtained from a coaxial X-band Gunn oscillator using distributed circuit elements. An analytical expression for the improved tuning range is presented and predictions for the improvement in an existing microstrip X-band oscillator using chip devices given.     

BP神经网络在元器件非工作可靠性参数预测中的应用

应用神经网络模型对某电子元器件在非工作状态时的可靠性性能参数时间序列进行预测,并与自回归模型预测结果进行了比较.检验结果表明,神经网络预测模型有较高的精度,具有很好的实用价值.     

Simulation of dynamics of the neural ensemble in an active wireless network

The problem of emulation of the Hopfield networks in active wireless networks is considered. A network with six elements communicating via radio channels is studied experimentally.     

动力学神经网络模型的构建及其稳定性研究

针对静态人工神经网络具有在反映系统动态行为时网络结构复杂、不能很好地反映系统动态性能的缺点,提出一种由带有积分器和可调反馈系数的神经元构成的新型动力学神经网络模型。该网络比以前的动态网络即递归网络或在此基础上改进的网络能更好地反映系统的动态性能,网络的结构更加简单,训练过程加快,从而使系统能够更好的运行。利用梯度下降法研究了该网络的权值调整算法,并通过李雅普诺夫稳定性判据讨论了这种新型动力学神经网络的稳定性条件。该网络研究为反映系统的动力学行为提供了更好的模型结构和理论算法,为神经网络的发展提供了新的研究方向。     

基于BP神经网络的模塑封电子器件优化设计

针对塑封SOT(小外形晶体菅)器件的使用失效案例,从芯片设计角度出发,提出一种优化设计方法,该方法利用误差反向传播神经网络(BPNN),结合主成分分析(PCA)、遗传算法(GAs)及均匀设计的针对非线性系统的优化设计,设计了该塑封SOT器件的尺寸参数。结合实验和有限元模拟分析,验证该优化设计结果的有效性。结果表明,优化设计的器件各关键界面点的最大应力约减少了70~180MPa,器件的界面层裂现象得到消除,提高了器件的可靠性。     

A stochastic network model of the interaction between cardiacrhythm and artificial pacemaker

In order to study heart-pacemaker interaction (HPI), a computer model of the cardiac conduction system has been developed which includes the effects of artificial pacemaker function and failure. The stochastic network model of cardiac conduction consists of five vertices, each representing a functional electrophysiologic element. Electrophysiologic multidimensional conditional probability functions determine the depolarization status of each vertex. The atrioventricular (AV) node is emulated using a mathematical model which includes the influence of past cycle lengths on AV nodal conduction time. Twenty-three classes of arrhythmias may be simulated and, for pacing simulation, one of 12 antibradycardia pacing modes may be chosen. Random effects of pacemaker malfunction including oversensing, undersensing, or failure-to-capture may be simulated through the use of probability distribution functions. This model should prove useful in the development of pacemaker algorithms, determining patient-specific pacemaker therapy, and predicting causes for apparent pacemaker malfunction. The model has been used in the development of an expert system to analyze paced ECGs for pacemaker function and malfunction     

防空兵群火力单位机动部署的Hopfield神经网络模型研究

针对防空兵群火力单位机动部署问题,引入了基于Hopfield神经网络的解决方法．建立了神经网络模型,给出了神经网络能量函数的表示方法及状态方程,并在Matlab中进行了仿真实验,实验表明该方法是可行的。     

基于群卷积神经网络的恶意域名检验方法

针对恶意域名检测中存在的随机性大、现实样本少的缺陷,导致深度学习模型训练易出现过拟合的问题,提出了一种基于群卷积神经网络的恶意域名检测方法。首先将域名转换为嵌入词向量表示,然后通过随机维度组合生成随机数据集并构建卷积神经网络组,鉴于Inception结构优势将其加入到网络中,最后针对数据集易出现的类间样本失衡问题,引入了类间平衡系数以抑制模型训练过拟合,提高模型泛化能力。实验结果表明,在采集的域名检测数据集上,所构建的模型能够有效实现恶意域名检测;经过参数优化,相比于浅层模型组合分类器与典型深度神经网络模型LSTM-CNN,群卷积神经网络对所构建的域名检测集检测准确率分别提升了4%、1%,达到98.9%。     

A neural network compensator for uncertainties of roboticsmanipulators

A neural network controller for trajectory control of robotic manipulators that is used not to internalize the inverse dynamic model of the controlled object but to compensate only the uncertainties of the robotic manipulator is presented. Its performance is compared with that of the conventional adaptive scheme. The results show the ability of the neural network controller to adapt to unstructured effects. A learning method for the neural network compensator with true teaching signals is shown. The tracking error of the robotic manipulator was greatly reduced when this controller was used     

Experimental studies on the electronic and mechanical tuning properties of resonant-cap IMPATT oscillators

A detailed experimental investigation has been made on the electronic tuning of resonant-cap IMPATT oscillators at the X-band. The investigation is extended to include mechanical tuning properties, i.e. studies on the dependence of the tuning properties on the resonant-cap diameter and the cap height. The study of electronic tuning gives an interesting insight regarding tuning behaviour for various ranges of d.c. bias current. From the study of these tuning properties the optimum size of the resonant-cap cavity for realizing maximum power output and electronic tunability for a given IMPATT diode can be obtained, which would be useful in the design of resonant-cap IMPATT oscillators.     

基于神经网络的雷达抗干扰方法综述

针对目前雷达抗干扰方法存在的局限性,提出了将神经网络技术应用于雷达抗干扰的研究思路。首先介绍了在雷达抗干扰领域应用比较成功的3种神经网络模型:BP网络、RBF网络和Kohonen网络;然后概括了神经网络应用于雷达抗干扰的3种主要途径:神经网络用于建模、用于特征提取和用于分类;最后对下一步的研究进行了总结与展望,分析了神经网络与其他技术相结合在雷达抗干扰中的应用前景。     

A CASCADED MODEL OF NEURAL NETWORK FOR PATTERN RECOGNITION

A cascaded model of neural network and its learning algorithm suitable for opticalimplementation are proposed.Computer simulations have shown that this model may successfullybe applied to an error-tolerance pattern recognitions of multiple 3-D targets with arbitrary spatialorientations.     

Hopfield神经网络的存贮容量

我们把理论推导与数值模拟相结合得出一个较好的误差函数近似解析式。应用该解析式分析了Ｈｏｐｆｉｅｌｄ神经网络绝对存同容量，得到了一更严格的结果。     

一种基于组合神经网络的HEVC帧间预测方法

高效视频编码(HEVC)作为最新视频编码标准,有着非常高的压缩效率,但是由于各种新技术的提出,其编码复杂度也大大提高。复杂度对视频编码有着重要意义,低复杂度编码的研究非常必要。利用神经网络进行HEVC的分区预测为低复杂度编码提供了有效的解决方案。文中提出了一种基于卷积神经网络(CNN)和长短期记忆网络(LSTM)的组合网络架构来对帧间分区进行预测的方法,利用自建数据库对网络进行训练;文中设计了一种预搜索模块来建立训练数据库,仿真结果表明,神经网络的精度可达87%,利用该网络架构进行帧间预测可以实现52%~71%的复杂度节省。     

Influence of the model for random birefringence on the differential group delay of periodically spun fibers

We consider the two Wai-Menyuk models of birefringence in periodically spun fibers, and we show that the differential group delay differs significantly for the two models when the spin period approaches or exceeds the fiber beat length. When the fiber correlation length is large, we explain this difference quantitatively, and we explain it qualitatively for any fiber correlation length.     

A neural network approach for compact cryogenic modelling of HEMTs

This paper reports on a procedure based on the use of artificial neural networks (ANN) to fully model the performance of advanced high electron mobility transistors (HEMT) operating down to cryogenic temperatures. By means of this procedure, we reproduce the DC behaviour and the scattering (S-) parameters of the device under test (DUT). The I-V curves and the S-parameters of the DUT have been compared with measurements, and a good agreement has been found for assessing the capability of the ANN structure to predict the full behaviour of the DUT. Furthermore, we have analysed in detail the performance of two typical parameters of HEMT's, namely the transconductance and the output conductance. Their values have been derived from measured data and have been compared with those obtained by the ANN approach. Both the simulated DC and RF performance have shown an accuracy degree adequate to model the device properties down to cryogenic temperatures.     

A quick neural network for computer vision of gray images

This paper presents a comparison between the traditional image processing method and the area vector concept as well as the new technique of artificial neural networks. Freeman chain coding is considered in the study, and the principle of segmentation may be based and implemented for further investigations resulting from the proposed work. The pattern recognition concept is analyzed and defined through the sigmoid function and the determination of the threshold of a gray image for an object. The block schemes for the given protocols are summarized in a single scheme for illustration and comparison purposes. The synthetic pictures are generated and investigated regarding the dependence of computer vision on the contents of the artificial neural network. The normalization technique is included to eliminate noise and zooming problems. The minimum computational time for image processing with the generated pictures is also determined. The rate of deflection in the computational time is recommended for sensing the minimum computational time according to the variation of the number of hidden units in the hidden layer. A three-layer neural network has been used. The study of gray binary imaging for color pictures is illustrated to save computational time and effort.     

A survey of neural network accelerator with software development environments

Recent years,the deep learning algorithm has been widely deployed from cloud servers to terminal units.And researchers proposed various neural network accelerators and software development environments.In this article,we have reviewed the representative neural network accelerators.As an entirety,the corresponding software stack must consider the hardware architecture of the specific accelerator to enhance the end-to-end performance.And we summarize the programming environments of neural network accelerators and optimizations in software stack.Finally,we comment the future trend of neural network accelerator and programming environments.     

一种基于神经网络的目标优先级确定方法

传感器管理的目的之一是将多传感器资源分配给多目标。分配时需要考虑的因素之一是目标的优先级别。本文分析了影响目标优先级的主要因素，提出了一种基于人工神经网络的目标优先级确定算法。仿真结果表明，该算法是精确而有效的。