北京地区CHINANET的现状与未来发展之我见

介绍了CHINANET的网络结构，分析了北京地区CHINANET网的现状，对CHINANET的发展、经营管理及安全问题提出了个人见解。     

ZXPCS无线市话(小灵通)系统的网络规划与优化

介绍了ZXPCS系统网络规划和优化的必要性、网络规划的方法和步骤、网络优化的目的和要求，以及网络优化中基站调整的原则等。     

A branch and bound network approach to the canonical constrained entropy problem

In this paper, we present a branch and bound algorithm for solving the constrained entropy mathematical programming problem. Unlike other methods for solving this problem, our method solves more general problems with inequality constraints. The advantage of the proposed technique is that the relaxed problem solved at each node is a singly constrained network problem. The disadvantage is that the relaxed problem has twice as many variables as the original problem. An application to regional planning is given, and an example problem is solved.     

Influence of chlorosulfonated polyethylene network structure on poly(acrylonitrile-styrene-acrylate)/poly(α-methylstyrene-acrylonitrile) blends: Mechanical properties,morphologies, glass transition behavior and heat resistance

Acrylonitrile-styrene-acrylate terpolymer and poly(α-methylstyrene-acrylonitrile) (ASA/α-MSAN, 25/75) with different loadings of chlorosulfonated polyethylene (CSM) were prepared via melt blending, with goals of toughening modification of ASA/α-MSAN blends and maintaining the heat resistance simultaneously. The results revealed CSM had excellent toughening effect at room temperature. At 0 °C, impact strength increased linearly with CSM content. However, toughening effect of CSM was undesirable at −30 °C. The temperature-dependent toughening efficiency of CSM was significantly related to its glass transition behavior. Scanning electron microscope analysis on cryo-fractured surfaces revealed the toughening mechanism was the formation of CSM toughening network in matrix, which was further confirmed by selective extraction tests. The formation of CSM network could lead to increased glass transition temperature of the blends at the low temperature region according to dynamic mechanical thermal analysis. Different from other toughening agents, CSM network uncompromised the heat resistance of ASA/α-MSAN blends.     

基于图神经网络的电子设计自动化技术研究进展

在摩尔定律的推动下,工艺节点在不断演进,集成电路设计复杂度也在不断增加,电子设计自动化(EDA)技术面临着来自运行时间与计算资源等诸多方面的挑战。为了缓解这些挑战,机器学习方法已被纳入EDA工具的设计流程中。与此同时,鉴于电路网表作为图形数据的本质,图神经网络(GNN)在EDA流程中的应用正变得越来越普遍,为复杂问题的建模以及最优问题的求解带来了新思路。该文首先对GNN与EDA技术的概念内涵进行了简要的概述,详细地梳理了GNN在高层次综合(HLS)、逻辑综合、布图规划与布局、布线、反向工程、硬件木马检测以及测试点插入等不同EDA设计流程中的主要作用,以及当前基于GNN的EDA技术的一些重要探索。以希望为集成电路设计自动化以及相关领域的研究人员提供参考,为我国先进集成电路产业的发展提供技术支持。     

小样本图像处理方法赋能的宽带频谱感知

针对强噪声环境下频谱感知方法计算复杂度高、难以获取大量标注样本、检测准确率低等问题,该文提出由图像去噪和图像分类思想驱动的频谱感知方法(IDCSS)。首先,对感知用户的接收信号进行时频变换,将无线电数值信号转换为图像。强噪声环境下感知用户接收信号图像与噪声图像相关度高,因此搭建生成对抗网络(GAN)来增加低信噪比下接收信号样本的数量,提高图像的质量。在生成器中,利用残差-长短时记忆网络取代生成网络U-Net结构中的跳跃连接,对图像进行去噪、提取感知用户接收信号图像的多尺度特征、建立基于熵的损失函数来构建网络的抗噪能力;在判决器中,设计适用无线电图像信号的多维度判决器来增强生成图像的质量、保留低信噪比感知用户信号的图像细节。最后利用分类器识别频谱占用状态。仿真结果表明,与现有频谱感知算法相比,所提算法具有较好的检测性能。     

Efficient routing on small complex networks without buffers

In this paper, we are exploring strategies for the reduction of the congestion in the complex networks. The nodes without buffers are considered, so, if the congestion occurs, the information packets will be dropped. The focus is on the efficient routing. The routing strategies are compared using two generic models, i.e., Barabàsi–Albert scale-free network and scale-free network on lattice, and the academic router networks of the Netherlands and France. We propose a dynamic deflection routing algorithm which automatically extends path of the packet before it arrives at congested node. The simulation results indicate that the dynamic routing strategy can further reduce the number of dropped packets in a combination with the efficient path routing proposed by Yan et al. [5].     

基于生成对抗网络的舰船辐射噪声分类方法研究

基于机器学习的舰船目标识别近年来已成为水声信号处理领域的一个重要研究方向,但水声目标信号的获取困难,样本量不足和不均衡的问题很容易导致目标分类模型的识别效果不佳。该文提出一种基于条件卷积生成对抗网络的船舶噪声数据分类方法,该方法利用生成对抗学习理论,生成相比于传统数据增强算法非线性特征更强,特征差异更丰富的伪DEMON调制谱数据来缓解训练样本量不足的问题。之后将传统生成对抗网络中的全连层输出替换成更善于解决小样本问题集成分类器,从而降低分类器对于数据量的依赖程度,进一步提高分类模型性能。最终由基于真实样本的实验结果表明,相比于传统数据增强算法和卷积生成对抗网络,该文方法能够更有效提高在样本不足条件下的模型的分类性能。     

网络优化中一个组合问题的探讨

文[4]提出了网络优化中若干有待解决的组合问题，本文围绕其中之一“减小直径问题”进行了探讨．设P（n，t）表示长为n的路径增加t条边后所得图直径的最小值，C（n，t）表示长为n的圈增加t条边后所得回直径的最小值．本文取得如下进展：1）给出P（n，2），P（n，3）及C（n，2）的精确值，并得出P（n，4）的一更精细的上界及一种更好的加边方式．上述结果均满足小极大度原则．2）在有极大度限制的条件下，分别对t为偶数和奇数给出了P（n，t）的上界．     

基于自动秩估计的黎曼优化矩阵补全算法及其在图像补全中的应用

矩阵补全(MC)作为压缩感知(CS)的推广,已广泛应用于不同领域。近年来,基于黎曼优化的MC算法因重构精度高、计算速度快的特点,引起了广泛关注。针对基于黎曼优化的MC算法需假设原矩阵秩固定已知,且随机选择迭代起点的特点,该文提出一种基于自动秩估计的黎曼优化MC算法。该算法通过优化包含秩正则项的目标函数,迭代获取秩估计值和预重构矩阵。在估计所得秩对应的矩阵空间上以预重构矩阵为迭代起点,利用基于黎曼流形的共轭梯度法进行矩阵补全,从而提高重构精度。实验结果表明,与几种经典的图像补全方法相比,该文算法图像重构精度显著提高。