A Graph with Adaptive AdjacencyMatrix for Relation Extraction

The relation is a semantic expression relevant to two named entities in a sentence.Since a sentence usually contains several named entities,it is essential to learn a structured sentence representation that encodes dependency information specific to the two named entities.In related work,graph convolutional neural networks are widely adopted to learn semantic dependencies,where a dependency tree initializes the adjacency matrix.However,this approach has two main issues.First,parsing a sentence heavily relies on external toolkits,which can be errorprone.Second,the dependency tree only encodes the syntactical structure of a sentence,which may not align with the relational semantic expression.In this paper,we propose an automatic graph learningmethod to autonomously learn a sentence’s structural information.Instead of using a fixed adjacency matrix initialized by a dependency tree,we introduce an Adaptive Adjacency Matrix to encode the semantic dependency between tokens.The elements of thismatrix are dynamically learned during the training process and optimized by task-relevant learning objectives,enabling the construction of task-relevant semantic dependencies within a sentence.Our model demonstrates superior performance on the TACRED and SemEval 2010 datasets,surpassing previous works by 1.3%and 0.8%,respectively.These experimental results show that our model excels in the relation extraction task,outperforming prior models.

Structural Synthesis of Compliant Metamorphic Mechanisms Based on Adjacency Matrix Operations

A compliant metamorphic mechanism attributes to a new type of metamorphic mechanisms evolved from rigid metamorphic mechanisms. The structural characteristics and representations of a compliant metamorphic mechanism are different from its rigid counterparts, so does the structural synthesis method. In order to carry out its structural synthesis, a constraint graph representation for topological structure of compliant metamorphic mechanisms is introduced, which can not only represent the structure of a compliant metamorphic mechanism, but also describe the characteristics of its links and kinematic pairs. An adjacency matrix representation of the link relationships in a compliant metamorphic mechanism is presented according to the constraint graph. Then, a method for structural synthesis of compliant metamorphic mechanisms is proposed based on the adjacency matrix operations. The operation rules and the operation procedures of adjacency matrices are described through synthesis of the initial configurations composed of s＋1 links from an s-link mechanism （the final configuration）. The method is demonstrated by synthesizing all the possible four-link compliant metamorphic mechanisms that can transform into a three-link mechanism through combining two of its links. Sixty-five adjacency matrices are obtained in the synthesis, each of which corresponds to a compliant metamorphic mechanism having four links. Therefore, the effectiveness of the method is validated by a specific compliant metamorphic mechanism corresponding to one of the sixty-five adjacency matrices. The structural synthesis method is put into practice as a fully compliant metamorphic hand is presented based on the synthesis results. The synthesis method has the advantages of simple operation rules, clear geometric meanings, ease of programming with matrix operation, and provides an effective method for structural synthesis of compliant metamorphic mechanisms and can be used in the design of new compliant metamorphic mechanisms.

Configuration Analysis of Metamorphic Mechanisms Based on Extended Adjacency Matrix Operations

The adjacency matrix operations,which connect with configuration transformation correspondingly,can be used for analysis of configuration transformation of metamorphic mechanisms and the corresponding algorithm can easily be simulated by computer.But the adjacency matrix based on monochrome topological graph is not suitable for the topological representation of mechanisms with multiple joints.The method of adjacency matrix operations has its own limitations for analysis of configuration transformation of metamorphic mechanisms because it can only be used in the topological representation of mechanisms with single joints.In order to overcome the drawback of the adjacency matrix,a kind of new matrix named as extended adjacency matrix is proposed to express topological structures of all mechanisms.The extended adjacency matrix is not only suitable for the topological representation of mechanisms with single joints,but also can be used in that of mechanisms with multiple joints.On this basis,a method of matrix operations based on the extended adjacency matrix is proposed to analyze the configuration transformation of metamorphic mechanisms.The method is not only suitable for configuration analysis of metamorphic mechanisms with single joints as well as metamorphic mechanisms with multiple joints.The method is evaluated by calculating two examples representing metamorphic mechanisms with single joint and multiple joints respectively.It can be concluded that the method is effective and correct for analysis of configuration transformation of all metamorphic mechanisms.The proposed method is simple and easy to be achieved by computer programming.It provides a basis for structural synthesis of all metamorphic mechanisms.

Drawing Weighted Directed Graph from It's Adjacency Matrix

This paper proposes an algorithm for building weighted directed graph, defmes the weighted directed relationship matrix of the graph, and describes algorithm implementation using this matrix. Based on this algorithm, an effective way for building and drawing weighted directed graphs is presented, forming a foundation for visual implementation of the algorithm in the graph theory.

Probability Distribution of Edge in Adjacent Matrix of Aviation Network of China and Algorithm of Searching Non-overlap Community Structure Based on Complex Network

In order to discover the probability distribution feature of edge in aviation network adjacent matrix of China and on the basis of this feature to establish an algorithm of searching non-overlap community structure in network to reveal the inner principle of complex network with the feature of small world in aspect of adjacent matrix and community structure,aviation network adjacent matrix of China was transformed according to the node rank and the matrix was arranged on the basis of ascending node rank with the center point as original point.Adjacent probability from the original point to extension around in approximate area was calculated.Through fitting probability distribution curve,power function of probability distribution of edge in adjacent matrix arranged by ascending node rank was found.According to the feature of adjacent probability distribution,deleting step by step with node rank ascending algorithm was set up to search non-overlap community structure in network and the flow chart of algorithm was given.A non-overlap community structure with 10 different scale communities in aviation network of China was found by the computer program written on the basis of this algorithm.

Using adjacency matrix to explore remarkable associations in big and small mineral data

Data exploration,usually the first step in data analysis,is a useful method to tackle challenges caused by big geoscience data.It conducts quick analysis of data,investigates the patterns,and generates/refines research questions to guide advanced statistics and machine learning algorithms.The background of this work is the open mineral data provided by several sources,and the focus is different types of associations in mineral properties and occurrences.Researchers in mineralogy have been applying different techniques for exploring such associations.Although the explored associations can lead to new scientific insights that contribute to crystallography,mineralogy,and geochemistry,the exploration process is often daunting due to the wide range and complexity of factors involved.In this study,our purpose is implementing a visualization tool based on the adjacency matrix for a variety of datasets and testing its utility for quick exploration of association patterns in mineral data.Algorithms,software packages,and use cases have been developed to process a variety of mineral data.The results demonstrate the efficiency of adjacency matrix in real-world usage.All the developed works of this study are open source and open access.

广义θ-图和广义梅花图φ的奇异性

利用奇异图的邻接矩阵行列式等于零的方法讨论广义θ-图和广义梅花图φ的奇异性,分别给出广义θ-图θ(a_(1),a_(2),…,a_(k))和广义梅花图φ(a_(1),a_(2),…,a_(k))是奇异图的充分必要条件,并计算这两类图中奇异图发生的概率值.

二部图性质的谱刻画

为了刻画二部图的性质,研究了图的邻接矩阵、邻接特征多项式、线图、关联矩阵、拉普拉斯矩阵、无符号拉普拉斯矩阵等。用图的谱性质刻画了二部图的特征,并得到了以下结论:二部图G的奇数阶谱矩为0,邻接谱在实数轴上关于原点对称,–2是线图l(G)的重数为m−n+1的特征值,拉普拉斯矩阵和无符号拉普拉斯矩阵有相同的谱,最小无符号拉普拉斯特征值等于0,最大拉普拉斯特征值等于最大无符号拉普拉斯特征值。

基于会话的独立邻域矩阵偏好交互推荐

通过合并输入和输出邻域矩阵可以使一些工作生成全局和局部偏好,并直接对这2个偏好建模来构建会话表示,从而实现改进。然而,一个会话的输入矩阵和输出矩阵并没有很强的相关性,它们的连接可能会为构建2个偏好引入噪声。其次,全局偏好和局部偏好可以相互促进,且邻域会话的协同信息可能有助于提高推荐性能。因此,一种基于会话的偏好交互推荐被提出,它来自独立的输入邻域矩阵和输出邻域矩阵框架,包括2个并行模块:输入会话表示编码器(ISE)和输出会话表示编码器(OSE)。ISE通过GNN和并行协同注意力机制对具有输入信息的会话表示进行建模。OSE通过GNN和并行协同注意力机制对具有输出信息的会话表示进行建模。最后,引入一种融合门控机制来平衡ISE和OSE产生的会话表示的重要性。结果表明,在Yoochoose和Diginetica数据集上,提出的模型明显优于其他先进的方法。

高斯混合模型与文本图卷积网络结合的虚假评论识别算法

针对文本图卷积网络(Text GCN)窗口边权阈值策略不足的问题,为了更精准地挖掘相关的词关联结构、提高预测精度,提出一种高斯混合模型(GMM)与Text GCN结合的虚假评论识别算法F-Text GCN。首先,利用GMM分离噪声边权分布的特性,提高虚假评论在训练数据上相对正常评论数不足的边信号强度;然后,考虑到信源的多样性,综合文档、词汇和评论以及非文本特征构造邻接矩阵;最后,通过Text GCN的谱分解提取邻接矩阵的虚假评论关联结构实施预测。根据国内某大型电商平台采集的126086条实际中文评论数据开展实证研究,实验结果表明,F-Text GCN识别虚假评论的F1值达到82.92%,与预训练表征模型BERT和文本卷积神经网络相比分别提升了10.46%和11.60%,相较于只使用评论文本信源的Text GCN模型F1值提升了2.94%;研究了高仿虚假评论的预测错误率,在支持向量机(SVM)作用后难识别的评论样本上尝试二次识别,F-Text GCN整体预测准确率可达94.71%,相较于Text GCN和SVM,在识别准确率上分别提升了2.91%和14.54%。研究发现,虚假评论的二阶图邻居结构显示出较强的干预消费者决策的词汇,这表明所提算法特别适用于提取用于虚假评论检测的长程词语搭配结构和全局句子特征模式变化的场景。

基于随机矩阵的变电站多源异常声音信号的定向方法

变电站的安全稳定运行至关重要,通过分析变电站中电力设备的可听声特征变化可以判断设备是否处于不正常运行状态,对异常声源进行主动定位可以快速确定故障区段,进而提高异常声音的检测效率。在实际应用中,由于准确性和实时性的要求常采用大阵列小快拍系统,造成当多个目标声源角度相近时,传统的声音定位方法不能对其有效分辨。根据随机矩阵理论提出基于均匀平面矩阵的声源定位方法,利用Stieltjes变换校正样本协方差矩阵的特征值和特征向量,然后通过空间谱的谱峰搜索对异常声源定位,可以有效提高分辨相邻声源的成功率。研究结果表明,所提方法可以使用更少的快拍数分辨相邻声源,且分辨成功率得到有效提升。

基于改进超邻接矩阵和熵权-TOPSIS的卫星互联网关键节点识别

“一星多用、多星组网、多网协同”思想的发展与应用为卫星互联网的关键节点识别带来了更多的挑战,也提出了更高的要求。针对卫星时序网络节点评估结果不准确的问题,考虑了不同时间片拓扑之间的耦合强度,提出了一种基于改进超邻接矩阵(supra-adjacency matrix,SAM)的卫星互联网时序网络模型。随后,综合卫星节点在网络中固有的拓扑特性和通信特性,选取特征向量中心性、介数中心性、节点紧密度、传输时延、传输速率和传输容量指标建立了节点重要度综合评估指标体系,在此基础上,基于熵权-逼近理想解排序法(technique for order preference by similarity to an ideal solution,TOPSIS)和时间权重矩阵设计了卫星互联网节点重要度评估方法。通过ARPANET和铱星星座进行仿真验证,实验结果证明了所提出的模型和方法能够准确地从局部和全局角度获得卫星节点重要度排序,并识别出潜在重要节点,对卫星互联网关键节点识别及抗毁性研究有一定的参考意义。

风洞试验模型曲面快速处理系统开发

目的结合风洞试验模型曲面处理方法与CATIA CAA二次开发技术,开发曲面快速处理系统,以解决目前风洞试验模型曲面处理自动化程度低、重叠与搭接曲面检查困难及孔洞填充效率低等问题,提高设计效率、规范设计流程。方法提出一种基于边界表示法的特征智能识别算法,该方法基于边界表示法提取曲面的拓扑边界,通过拓扑分析和构建加权的邻接矩阵计算目标特征,实现待处理目标的快速识别和提取。基于该算法与CAA接口开发风洞试验模型曲面快速处理系统,以交互式形式进行发布,以某导弹风洞试验模型的曲面处理作为应用实例,验证该算法和曲面快速处理系统的适用性。结果通过实例演示,验证了该系统在重叠与搭接曲面检查、复杂孔洞填充以及批量化模型缩放的处理方面效果显著,相比于传统设计时手动曲面处理,效率提高了30倍。结论基于边界表示法的特征智能识别算法可以快速进行目标特征的识别,曲面快速处理系统可以有效地提高设计效率,实现了风洞试验模型曲面处理的快速化和智能化。

融合改进图卷积的跨模态检索

针对现有跨模态检索在公共子空间度量时难以充分挖掘模态内局部一致性的问题,提出了一种融合改进图卷积的跨模态检索方法。为了提升各模态内的局部一致性,以单个完整样本为节点构建模态图,充分挖掘特征间的交互信息;为了解决图卷积网络只能做浅层学习的问题,采用在每一层图卷积添加初始残差链接和权重恒等映射的方法来缓解此现象;为了通过高低阶邻居信息共同更新中心节点特征,提出减少邻居节点、增加图卷积网络层数的改进;为了学习高度局部一致且语义一致的公共表征,共享公共表征学习层权重,并联合优化公共子空间中模态内的语义约束和模态间的模态不变约束。实验结果表明,在Wikipedia和Pascal sentence这两个跨模态数据集上,不同检索任务的平均mAP值比11种现有方法分别提升了2.2%~42.1%和3.0%~54.0%。

多项正则化约束的伪标签传播优化脑电信号聚类

作为一种非侵入式分析载体,脑电信号目前被广泛应用于脑-机接口、医疗辅助诊断及康复领域,但这些应用通常依赖需要完整标签的有监督分析技术,如分类.随着无标签脑电信号的与日俱增,现有的有监督方法不能有效解决无标签脑电信号分析问题,也在一定程度上限制了无标签脑电信号这类新型数据的应用拓展.为了解决无标签脑电信号的无监督分析问题,提出了一种基于多项正则化约束的伪标签传播优化聚类模型.该模型通过同时优化学习伪标签传播矩阵、脑电信号相似度邻接矩阵、标签分类器的方式实现聚类.将提出的脑电信号聚类模型转化为一个多目标优化问题,并提出了一种基于梯度下降策略的聚类算法EEGapc(electroencephalogram clustering with pseudo label propagation).该算法不仅充分考虑了脑电信号之间的相关性及脑电信号间的信息传递,还能快速收敛到局部最优.在14个真实脑电信号数据集上的实验结果表明,提出的EEGapc脑电信号聚类算法比现有的8种聚类算法性能更好,且在平均NMI(normalized mutual information),ARI(adjusted rand index),F-score,kappa这4个指标上,EEGapc与现有的8种聚类算法相比,分别至少提升了86.88%,58.01%,6.29%,61.17%.

基于时间感知图卷积网络的船舶航迹预测

针对船舶航迹数据非等距时间间隔的特点,提出一种基于时间感知图卷积网络的船舶航迹预测方法。该方法将船舶的时序航迹数据转换为图结构,结合改进的高斯核函数,以图节点间的时间距离和空间欧式距离构建时空带权邻接矩阵,显式建模非等距时间间隔采样点间的时空关联强度,以此捕捉船舶航迹数据的时空动态关系;通过图卷积网络和自注意力机制,依据时空带权邻接矩阵,提取图中节点的时空特征并加权,使模型聚焦于船舶航迹的关键时空信息。在船舶自动识别系统采集的公开数据上进行的仿真实验结果显示,与基于循环神经网络的方法相比,本文方法预测精度有显著提升。

混合树和混合圈的奇异性

为了刻画所有非奇异混合树和一些混合圈的奇异性,利用无向图或有向图的一些结论和线性代数的知识描述了混合图和有向图的秩,并得到以下结论:一个混合树是非奇异的当且仅当它具有完美匹配,且完美匹配中的每条边都是无向的;只含有一条无向边的n阶混合圈非奇异当且仅当n-1条弧方向一致;只含有一条有向边的n阶混合圈非奇异当且仅当n≠0(mod4);一个弧边交错的n阶混合圈非奇异当且仅当n=2(mod4)。

基于知识共享的遮挡人体姿态估计网络

现有人体姿态估计方法处理遮挡情况时性能较差,为此提出新的估计网络,包含遮挡区域强化卷积网络(OCNN)和遮挡特征补偿图卷积网络(OGCN).设计高低阶特征匹配注意力以强化遮挡区域特征,由OCNN提取高适配权重,通过少量遮挡数据的方式实现遮挡部位的强化检测.由OGCN消除障碍物特征,通过强化关键点共有及专有属性的方式补偿节点特征;进行邻接矩阵重要性加权以改善遮挡部位特征质量,提升检测精度.所提网络在数据集COCO2017、COCO-Wholebody、CrowdPose上的检测精度分别为78.5%、67.1%、77.8%,优于对比算法.在自建遮挡数据集上所提网络节约了75%的训练数据使用.

高技术产业创新活动对区域经济发展的驱动作用及其影响机制

厘清区域高技术产业如何赋能创新活动以助推地区经济发展,是提升区域发展整体效能、有效应对全球科技竞争的战略性问题。针对目前研发投资驱动的自主研发创新和外国直接投资驱动的跨国创新引入的作用机制及空间外溢的差异未明确,以及高技术产业技术市场化水平对不同创新活动经济效应的影响有待发掘的现状,将高技术产业科技密集度、技术市场化纳入创新驱动经济发展的研究框架,构建地理邻接矩阵和地理距离矩阵两种空间权重矩阵,以中国29个省份为研究对象,基于2011—2021年国家有关统计年鉴中的省级面板数据,运用动态空间杜宾模型实证分析自主研发创新、跨国创新引入对当地/邻近区域经济发展影响机制和空间外溢的差异性,探讨在高技术产业科技密集度及技术市场化水平的作用下,自主研发创新及跨国创新引入对经济发展的不同影响,并进一步分类对比分析高技术产业技术市场化水平、不同创新活动与区域经济发展的内在关系。结果表明:(1)区域创新活动水平与经济发展具有显著的空间依赖性,自主研发创新对经济发展的积极作用在长期内效果更明显,跨国创新引入对经济发展的积极作用在短期内效果更明显;(2)高技术产业科技密集度对当地的自主研发创新/跨国创新引入与经济发展之间的关系具有显著正向调节作用,且这一作用具有持久性;(3)随着高技术产业中技术市场化水平的提升,自主研发创新、跨国创新引入对区域经济发展的积极关系分别存在先下降后上升、先上升后下降的非线性变化,但这种正向作用整体力度呈现减弱趋势;(4)在科技密集区/欠密集区,随着高技术产业技术市场化水平的变化,不同创新活动对区域经济发展非线性影响的变化趋势虽然相同,但总体作用力度有所差异。最后从提高高技术产业的科技密集度和规范高新技术产业的技术市场化体系两方面提出相关对策建议。