DHSEGATs:distance and hop-wise structures encoding enhanced graph attention networks

Numerous works prove that existing neighbor-averaging graph neural networks(GNNs)cannot efficiently catch structure features,and many works show that injecting structure,distance,position,or spatial features can significantly improve the performance of GNNs,however,injecting high-level structure and distance into GNNs is an intuitive but untouched idea.This work sheds light on this issue and proposes a scheme to enhance graph attention networks(GATs)by encoding distance and hop-wise structure statistics.Firstly,the hop-wise structure and distributional distance information are extracted based on several hop-wise ego-nets of every target node.Secondly,the derived structure information,distance information,and intrinsic features are encoded into the same vector space and then added together to get initial embedding vectors.Thirdly,the derived embedding vectors are fed into GATs,such as GAT and adaptive graph diffusion network(AGDN)to get the soft labels.Fourthly,the soft labels are fed into correct and smooth(C&S)to conduct label propagation and get final predictions.Experiments show that the distance and hop-wise structures encoding enhanced graph attention networks(DHSEGATs)achieve a competitive result.

GGC:Gray-Granger Causality Method for Sensor Correlation Network Structure Mining on High-Speed Train

Vehicle information on high-speed trains can not only determine whether the various parts of the train are working normally,but also predict the train’s future operating status.How to obtain valuable information from massive vehicle data is a difficult point.First,we divide the vehicle data of a high-speed train into 13 subsystem datasets,according to the functions of the collection components.Then,according to the gray theory and the Granger causality test,we propose the Gray-Granger Causality(GGC)model,which can construct a vehicle information network on the basis of the correlation between the collection components.By using the complex network theory to mine vehicle information and its subsystem networks,we find that the vehicle information network and its subsystem networks have the characteristics of a scale-free network.In addition,the vehicle information network is weak against attacks,but the subsystem network is closely connected and strong against attacks.

Drivers for Inter-city Innovation Networks Across Chinese Cities:Modelling Physical Versus Intangible Effects

Cross-region innovation is widely recognized as an important source of the long-term regional innovation capacity.In the recent past,a growing number of studies has investigated the network structure and mechanisms of cross-region innovation collaboration in various contexts.However,existing research mainly focuses on physical effects,such as geographical distance and high-speed railway connections.These studies ignore the intangible drivers in a changing environment,the more digitalized economy and the increasingly solidified innovation network structure.Thus,the focus of this study is on estimating determinants of innovation networks,especially on intangible drivers,which have been largely neglected so far.Using city-level data of Chinese patents(excluding Hong Kong,Macao,and Taiwan Province of China),we trace innovation networks across Chinese cities over a long period of time.By integrating a measure on Information and Communications Technology(ICT)development gap and network structural effects into the general proximity framework,this paper explores the changing mechanisms of Chinese innovation networks from a new perspective.The results show that the structure of cross-region innovation networks has changed in China.As mechanisms behind this development,the results confirm the increasingly important role of intangible drivers in Chinese inter-city innovation collaboration when controlling for effects of physical proximity,such as geographical distance.Since digitalization and coordinated development are the mainstream trends in China and other developing countries,these countries'inter-city innovation collaboration patterns will witness dramatic changes under the influence of intangible drivers.

Multiscale Characteristics and Connection Mechanisms of Attraction Networks:A Trajectory Data Mining Approach Leveraging Geotagged Data

Urban tourism is considered a complex system,and multiscale exploration of the organizational patterns of attraction networks has become a topical issue in urban tourism,so exploring the multiscale characteristics and connection mechanisms of attraction networks is important for understanding the linkages between attractions and even the future destination planning.This paper uses geotagging data to compare the links between attractions in Beijing,China during four different periods:the pre-Olympic period(2004–2007),the Olympic Games and subsequent‘heat period’(2008–2013),the post-Olympic period(2014–2019),and the COVID-19(Corona Virus Disease 2019)pandemic period(2020–2021).The aim is to better understand the evolution and patterns of attraction networks at different scales in Beijing and to provide insights for tourism planning in the destination.The results show that the macro,meso-,and microscales network characteristics of attraction networks have inherent logical relationships that can explain the commonalities and differences in the development process of tourism networks.The macroscale attraction network degree Matthew effect is significant in the four different periods and exhibits a morphological monocentric structure,suggesting that new entrants are more likely to be associated with attractions that already have high value.The mesoscale links attractions according to the common purpose of tourists,and the results of the community segmentation of the attraction networks in the four different periods suggest that the functional polycentric structure describes their clustering effect,and the weak links between clusters result from attractions bound by incomplete information and distance,and the functional polycentric structure with a generally more efficient network of clusters.The pattern structure at the microscale reveals the topological transformation relationship of the regional collaboration pattern,and the attraction network structure in the four different periods has a very similar importance profile structure suggesting that the attraction network has the same construction rules and evolution mechanism,which aids in understanding the attraction network pattern at both macro and micro scales.Important approaches and practical implications for planners and managers are presented.

Modular Epidemic Spreading in Small-World Networks

We study the epidemic spreading of the susceptible-infected-susceptible model on small-world networks with modular structure. It is found that the epidemic threshold increases linearly with the modular strength. Furthermore, the modular structure may influence the infected density in the steady state and the spreading velocity at the beginning of propagation. Practically, the propagation can be hindered by strengthening the modular structure in the view of network topology. In addition, to reduce the probability of reconnection between modules may also help to control the propagation.

Topological phase transition in network spreading

This paper investigates information spreading from the perspective of topological phase transition.Firstly,a new hybrid network is constructed based on the small-world networks and scale-free networks.Secondly,the attention mechanism of online users in information spreading is studied from four aspects:social distance,individual influence,content richness,and individual activity,and a dynamic evolution model of connecting with spreading is designed.Eventually,numerical simulations are conducted in three types of networks to verify the validity of the proposed dynamic evolution model.The simulation results show that topological structure and node influence in different networks have undergone phase transition,which is consistent with the phenomenon that followers and individual influence in real social networks experience phase transition within a short period.The infection density of networks with the dynamic evolution rule changes faster and reaches higher values than that of networks without the dynamic evolution rule.Furthermore,the simulation results are compared with the real data,which shows that the infection density curve of the hybrid networks is closer to that of the real data than that of the small-world networks and scale-free networks,verifying the validity of the model proposed in this paper.

Virtual Organization Information Systems and its Reengineering for Agile Manufacturing

It is now possible to organize globally while worki ng locally: Information technologies like e-mail, the Internet, and video confere ncing to the desktop permit tight coordination of geographically dispersed worke rs across time zones and cultures. Companies are not limited to physical locatio ns for providing products and services. Networked information systems are allowi ng companies to coordinate their geographically distributed capabilities as virt ual organizations. In order for organizations to succeed, they must be able to r espond with agility in a geographically dispersed environment. The core for a vi rtual organization to increase the utilization rate of resources to the maxi mum and to make full use of the transient market opportunities lies in how to br ing the potential of information technology into play. Based on the philosophy o f agile manufacturing, this paper analyses the basic concepts and connotation of Virtual Organization Information systems(VOIS). VOIS is an information system composed of some independent information subsystems that are autonomous, collab orative and belong to umpty organizations respectively. VOIS support the operati on of virtual organization, and automate the information flow across organizatio nal boundaries. Such systems have capabilities as rapid construction, quick oper ation, and agile reengineering and swift adaptability. Differences between VOIS and traditional enterprise information systems are analyzed. On the basis of ana lyzing the structure of VOIS, an abstract hierarchical structure of VOIS is prop osed using object-oriented method. Reengineering ideology and methods of VOIS a re also explored in light of the three layers structure.

Significance of the astrocyte domain organization for qualitative information structuring in the brain

Astrocytes, the dominant glial cell type, modulate synaptic information transmission. Each astrocyte is organized in non-overlapping domains. Here, a formally based model of the possible significance of astrocyte domain organization is proposed. It is hypothesized that each astrocyte contacting n neurons with m synapses via its processes generates dynamic domains of synaptic interactions based on qualitative criteria so that it exerts a structuring of neuronal information processing. The formalism (morpho-grammatics) describes the combinatorics of the various astrocytic receptor types for occupancy with cognate neurotransmitters. Astrocytic processes are able both to contact synapses and retract from them. Rhythmic oscillations of the astrocyte may program the domain organization, where clock genes may play a role in rhythm generation. For the interpretation of a domain organization a player of a string instrument is used as a paradigm. Since astrocytes form networks (syncytia), the interactions between astrocyte domains may be comparable to the improvisations in a jazz ensemble. Given the fact of a high combinational complexity of an astrocyte domain organization, which is formally demonstrable, and an uncomputable complexity of a network of astrocyte domains, the model proposed may not be testable in biological brains, but robotics could be a real alternative.

Self-Organizing Genetic Algorithm Based Method for Constructing Bayesian Networks from Databases

The typical characteristic of the topology of Bayesian networks (BNs) is the interdependence among different nodes (variables), which makes it impossible to optimize one variable independently of others, and the learning of BNs structures by general genetic algorithms is liable to converge to local extremum. To resolve efficiently this problem, a self-organizing genetic algorithm (SGA) based method for constructing BNs from databases is presented. This method makes use of a self-organizing mechanism to develop a genetic algorithm that extended the crossover operator from one to two, providing mutual competition between them, even adjusting the numbers of parents in recombination (crossover/recomposition) schemes. With the K2 algorithm, this method also optimizes the genetic operators, and utilizes adequately the domain knowledge. As a result, with this method it is able to find a global optimum of the topology of BNs, avoiding premature convergence to local extremum. The experimental results proved to be and the convergence of the SGA was discussed.

信息网络地位与产业结构升级

数字技术发展促使信息流动打破时空局限,形成信息网络并深刻影响经济主体效率。以往研究通过交通基础设施、企业生产等形成的城市网络,研究信息在城市之间流动对产业结构的影响,但这并非衡量城市信息网络的最直接方式。本文基于2011—2017年中国城市之间的日频百度搜索数据,构建了城市间信息网络,并用信息联系强度、信息多样性刻画信息网络地位特征,并讨论其对产业结构升级的影响。结果发现,信息联系强度和信息多样性促进了城市产业结构升级,且仅在中西部地区显著;省间信息网络作用大于省内;促进知识溢出、带动城市创业是信息网络产生影响的重要渠道。本文为中国构建网络空间命运共同体提供了经验依据,也为数字经济时代城市提升产业结构提供了可行路径。

面向任务的载人深潜人机交互信息重要度评估

为更好地识别复杂任务环境下深海载人潜水器驾驶舱人机交互信息重要度差异,提出一种基于复杂网络的人机交互信息重要度评估方法。以悬停采样任务为研究对象,运用决策阶梯模型及任务-网络建模技术识别复杂任务相关人机交互信息元及逻辑关联,构建交互信息复杂网络模型。基于网络拓扑特征参数,从节点间逻辑影响关系的角度提出节点全局及局部影响效应评估指标,结合解释结构模型层级权重指标对交互信息节点重要度综合值进行计算,得出悬停采样任务人机交互信息重要度排序。将排序结果与其他4种算法进行对比,在证明计算有效的基础上,邀请潜航员结合实际任务流程对交互信息进行主观重要度评估。结果显示,研究所得重要度排序结果与潜航员主观排序结果呈强相关,且对重要交互信息元的识别结果与潜航员主观感知相符合,验证了所提方法识别复杂任务人机交互信息重要度差异的有效性。

基于结构功能交叉神经网络的多模态医学图像融合

针对多模态医学图像融合中存在纹理细节模糊和对比度低的问题,提出了一种结构功能交叉神经网络的多模态医学图像融合方法。首先,根据医学图像的结构信息和功能信息设计了结构功能交叉神经网络模型,不仅有效地提取解剖学和功能学医学图像的结构信息和功能信息,而且能够实现这两种信息之间的交互,从而很好地提取医学图像的纹理细节信息。其次,利用交叉网络通道和空间特征变化构造了一种新的注意力机制,通过不断调整结构信息和功能信息权重来融合图像,提高了融合图像的对比度和轮廓信息。最后,设计了一个从融合图像到源图像的分解过程,由于分解图像的质量直接取决于融合结果,因此分解过程可以使融合图像包含更多的细节信息。通过与近年来提出的7种高水平方法相比,本文方法的AG,EN,SF,MI,QAB/F和CC客观评价指标分别平均提高了22.87%,19.64%,23.02%,12.70%,6.79%,30.35%,说明本文方法能够获得纹理细节更清晰、对比度更好的融合结果,在主观视觉和客观指标上都优于其他对比算法。

氢化改性钒基催化剂降解氯苯的机理研究

为了在低温下高效、稳定地降解挥发性氯代有机污染物(CVOCs),以偏钒酸铵为VO_(x)前驱体制备2D纳米片与3D纳米结构的互穿多孔网络结构的L-V_(2)O_(5)催化剂,并通过氢化技术对催化剂中部分V^(5+)进行还原(H-V_(2)O_(5))。结果表明,该氢化技术降低了催化剂中部分V^(5+)的化学价态,增加了V—O(Ⅱ)和V—O(Ⅳ)上的氧空位数量和催化剂表面活性氧数量。在275℃时,相较于L-V_(2)O_(5),H-V_(2)O_(5)对氯苯催化降解的效率提高了约40%。

基于多孔卷积神经网络的图像空间结构信息细节表征

针对传统图像空间结构信息表征方法存在细节表征模糊度较高、信息训练损失较高等问题,提出一种新的基于多孔卷积神经网络的图像空间结构信息细节表征方法。该方法通过图像空间结构信息细节相似性度量,并以图像的形状、颜色和纹理特征对图像空间结构信息细节进行编码,再去除图像冗余信息,利用多孔卷积神经网络对图像空间结构的深度信息进行融合,从而完成图像空间结构信息的细节表征。实验结果表明,基于多孔卷积神经网络的图像空间结构信息细节表征方法在模糊度、训练损失、图像相似性等方面都比传统的3种方法优越,能够清晰地表征图像空间结构信息。

基于SEM和神经网络的移动社交电商用户隐私信息披露意愿研究

在移动社交电商情境中,用户的交互方式发生改变,个人隐私信息安全问题凸显,上述变化将会对用户的隐私信息披露意愿产生重要影响。为了探究移动社交电商用户隐私信息披露意愿的影响因素,本文基于S-O-R理论建立结构方程模型,旨在探究影响用户隐私信息披露意愿的影响因素及其作用机理。结果表明隐私政策有效性、隐私设置有效性、信息交互、情感交互等因素均正向显著影响用户隐私信息披露意愿。在此基础上,基于平台特征和用户感知收益构建了GA-BP神经网络预测模型,并针对不同预测结果提出具体应对措施。最终,文章从隐私信息安全保障和平台个性化服务等方面提出促进移动社交电商平台可持续发展的建议。

基于邻域信息的复杂网络节点重要性评估方法设计

在复杂网络中准确识别影响力节点,对网络管理和网络安全至关重要。局部中心性方法简明易用,但忽略了邻居节点间的拓扑关系,不能提供全局最优结果。提出了一种关联节点连边关系和拓扑结构的重要节点评估方法,该方法首先综合应用H指数和信息熵对节点进行评估,在此基础上代入节点的结构洞特征,即在关注节点自身质量和邻居节点信息量的同时,考虑了其“桥接”属性。采用疾病传播模型对算法进行验证,应用肯德尔(Kendall)相关系数、互补累积分布函数以及传播影响力来验证算法的有效性与适用性。在6个真实网络数据集上的仿真结果表明,在识别和排序网络中的关键节点上,所提方法比传统中心性方法更准确。

高校全光校园网建设项目的研究与应用——以沈阳建筑大学为例

随着高校师生对信息化使用需求的断崖式增长,校园网越来越受到师生们的青睐。校园网可以为师生提供优质的信息,从而满足其办公和学习的需求。为进一步促进高校信息化网络建设,针对传统高校校园网存在的问题,发挥以太全光网技术的优势,探究以极简以太全光网技术为基础的全光校园网建设。以沈阳建筑大学全光校园网建设为例,根据师生的网络利用需求,构建全光校园网建设项目,绘制校园核心区域网络拓扑结构图,设计全光网建设项目内容,展示全光宿舍网建设项目的应用效果。

基于潜层关系增强的实体和关系联合抽取

为充分发掘文本序列中潜层语义关系信息,提出一种实体和关系联合抽取的潜层关系增强模型SREM(text subtext relationship enhancement model)。在潜层关系表示层利用结构化对齐的方式,获取并保持文本序列中的语义信息结构。在融合注意力机制的关系网络层中对数据进行建模,提高模型对文本词汇间关系信息的捕获能力。结合注意力机制获取细粒度语义信息,对上下文信息进行选择过滤。实验结果表明,在数据集NYT和WebNLG上取得的F1值分别为92.40%和92.52%,验证了模型的有效性。

科技文献的多层次结构功能识别

实现科技文献结构功能的自动识别有助于提升细粒度信息检索、关键词抽取、引文分析等任务的效率。针对当前结构功能识别研究面临的文本内部依赖关系表达能力较弱、模型泛化迁移能力不足等问题,本研究利用图卷积神经网络捕捉单词节点间存在的固有依赖信息和拓扑结构,提升模型对科技文本建模表达能力,同时,还引入对抗学习思想,提升结构功能识别模型的泛化能力。选取ScienceDirect数据集,考察多种模型方法对章节标题、章节内容、章节段落三个不同层次的结构功能的识别效果,并在PubMed-20k的医学摘要结构功能数据集上进一步测试多种模型的跨领域迁移能力。研究结果表明,在章节标题层次,BERT+GCN的识别效果最佳,F1值达到了88%,比基线模型提升3%;在章节内容层次,BERT+GAN的识别效果最佳,F1值达到了76%,比基线模型提升了3%;在章节段落层次,F1值达到了68%。BERT+GCN的跨领域迁移能力相比其他模型更优,在跨领域数据上取得了90%的F1值。

用于多器官分割的多尺度聚合网络研究

多器官分割在病理分析、手术方案制定以及临床诊断上都具有重要的应用价值.但是,一些器官形变较大、尺寸较小且组织边缘模糊,导致分割效果较差.为了解决该问题,提出了一种用于多器官分割的多尺度聚合网络(MSANet).MSANet有两个优势:首先,设计了多尺度边界提取模块,使用多尺度卷积核提取多个特征图,将不同尺度的特征图互相结合,从而聚合全局上下文信息,并提取不同器官的边界和细节信息;其次,设计了聚焦式注意力模块,通过学习的注意力权重来调节特征图的重要性,从而聚焦感兴趣的多器官区域并捕捉不同器官的关键特征,进一步提高分割性能.在两个公开数据集CHAOS和MS-CMRSeg上进行了大量实验.实验结果表明:MSANet在两个数据集上的分割效果均优于当前主流的多器官分割方法,显著提高了多器官分割精度.