Research on Node Classification Based on Joint Weighted Node Vectors

Node of network has lots of information, such as topology, text and label information. Therefore, node classification is an open issue. Recently, one vector of node is directly connected at the end of another vector. However, this method actually obtains the performance by extending dimensions and considering that the text and structural information are one-to-one, which is obviously unreasonable. Regarding this issue, a method by weighting vectors is proposed in this paper. Three methods, negative logarithm, modulus and sigmoid function are used to weight-trained vectors, then recombine the weighted vectors and put them into the SVM classifier for evaluation output. By comparing three different weighting methods, the results showed that using negative logarithm weighting achieved better results than the other two using modulus and sigmoid function weighting, and was superior to directly concatenating vectors in the same dimension.

Diagnostic Power of Diffusion-weighted Magnetic Resonance Imaging for the Presence of Lymph Node Metastasis:A Meta-analysis

Present work was designed to quantitatively evaluate the performance of diffusion-weighted magnetic resonance imaging（DWI） in the diagnosis of the presence of metastasis in lymph nodes（LNs）. Eligible studies were identified from systematical Pub Med and EMBASE searches. Data were extracted. Meta-analyses were performed to generate pooled sensitivity and specificity on the basis of per-node, per-lesion and per-patient, respectively. Fourteen publications（2458 LNs, 404 lesions and 334 patients） were eligible. Per-node basis demonstrated the pooled sensitivity and specificity was 0.82（P〈0.0001） and 0.90（P〈0.0001）, respectively. Per-lesion basis illustrated the pooled sensitivity and specificity was 0.73（P=0.0036） and 0.85（P〈0.0001）, respectively. Per-patient basis indicated the pooled sensitivity and specificity was 0.67（P=0.0909） and 0.86（P〈0.0001）, respectively. In conclusion, DWI has rather a negative predictive value for the diagnosis of LN metastasis presence. The difference of the mean apparent diffusion coefficients between benign and malignant LNs is not yet stable. Therefore, the DWI technique has to be further improved.

Utility of diffusion-weighted imaging in the diagnosis of inguinal lymph node metastasis with malignant melanoma

Malignant melanoma is a malignancy of pigmentproducing cells(melanocytes) located predominantly in the skin. Nodal metastases are an adverse prognostic factor compromising long term patient survival. Therefore, accurate detection of regional nodal metastases is required for optimization of treatment. Computed tomography(CT) and magnetic resonance imaging(MRI) remain the primary imaging modalities for regional staging of malignant melanoma. However, both modalities rely on size-related and morphological criteria to differentiate between benign and malignant lymph nodes, decreasing the sensitivity for detection of small metastases. Surgery is the primary mode of therapy for localized cutaneous melanoma. Patients should be followed up for metastases after surgical removal. We report here a case of inguinal lymph node enlargement with a genital vesicular lesion with a history of surgery for malignant melanoma on her thigh two years ago. CT and diffusion weighted-MRI(DW-MRI) were applied for the lymph node identification. DW-MRI revealed malignant lymph nodes due to malignant melanoma metastases correlation with pathological findings.

WSN中基于Stateless Weight的移动代理路由设计

为了使移动代理的路由不再过分依赖于网络结构,提出了一种应用在无线传感器网络(WSN)中的移动代理的路由设计(SWR-MA)。在SWR-MA中,引入了一个与节点位置有关的参数Weight,通过比较Weight值,移动代理可以自主的确定自己的路径。SWR-MA可以应用在拓扑改变的网络中,比如带有移动Sink节点的网络中。最后,对该路由算法进行了仿真评价。

Using Mathematical Models in Decision Making Methodologies to Find Key Nodes in the Noordin Dark Network

A Dark Network is a network that cannot be accessed through tradition means. Once uncovered, to any degree, dark network analysis can be accomplished using the SNA software. The output of SNA software includes many measures and metrics. For each of these measures and metric, the output in ORA additionally provides the ability to obtain a rank ordering of the nodes in terms of these measures. We might use this information in decision making concerning best methods to disrupt or deceive a given dark network. In the Noordin Dark network, different nodes were identified as key nodes based upon the metric used. Our goal in this paper is to use methodologies to identify the key players or nodes in a Dark Network in a similar manner as we previously proposed in social networks. We apply two multi-attribute decision making methods, a hybrid AHP & TOPSIS and an average weighted ranks scheme, to analyze these outputs to find the most influential nodes as a function of the decision makers’ inputs. We compare these methods by illustration using the Noordin Dark Network with seventy-nine nodes. We discuss sensitivity analysis that is applied to the criteria weights in order to measure the change in the ranking of the nodes.

Ad Hoc On-Demand Multipath Distance Vector Routing Protocol Based on Node State

To improve the performance of Ad hoc on-demand multipath distance vector (AOMDV) protocol, we proposed NS-AOMDV which is short for “AOMDV based on node state”. In NS-AOMDV, we introduce node state to improve AOMDV’s performance in selecting main path. In route discovery process, the routing update rule calculates the node weight of each path and sorts the path weight by descending value in route list, and we choose the path which has the largest path weight for data transmission. NS-AOMDV also uses the technology of route request (RREQ) packet delay forwarding and energy threshold to ease network congestion, limit the RREQ broadcast storm, and avoid low energy nodes to participate in the establishment of the path. The results of simulation show that NS-AOMDV can effectively improve the networks’packets delivery rate, throughput and normalized routing overhead in the situation of dynamic network topology and heavy load.

无线传感网络拥塞控制中节点选择强制博弈方法

无线传感网络中节点数据流量较大,分布范围较广,节点选择寻优计算复杂,导致浪费大量算力在节点选择上,拥塞控制效果不佳。提出一种无线传感网络拥塞控制中节点选择强制博弈方法,计算单条网络链路传输成本,依据传输数据耗费能量,计算传输节点可用能量比值,在传输变量权重值基础上,求得传输节点实际拥堵指数,完成节点选择。计算网络连接层的数据包丢弃概率,确定网络拥塞程度,汇聚节点数据流建立节点选择强制博弈模型,明确传感网络稳态传输条件,利用流量正态分布算法,在节点选择博弈中做出强制选择。经仿真分析证明:所提方法分组递交率保持在83%以上,丢包数量保持在100 pkt/s之内,网络传输节点端到端的延迟在0.3 s以内。

The Next-Hop Node Selection Based GPSR in Vehicular Ad Hoc Networks

Vehicular Ad hoc Networks (VANETs) which is a special form of Mobile Ad hoc Networks (MANETs) has promising application prospects in the future. Due to the rapid changing of topology structure, how to find a route which can guarantee Quality of Service (QoS) is an important issue in VANETs. This paper presents an improved Greedy Perimeter Stateless Routing (GPSR) protocol based on our proposed next-hop node selection mechanism. Firstly, we define the link reliability in two cases which take the movement direction angle between two vehicles into consideration. Then we propose a next-hop node selection mechanism based on a weighted function which consists of link reliability between the sender node and next-hop candidate node, distance between next-hop candidate node and the destination, movement direction angle of next-hop candidate node. At last, an improved GPSR protocol is proposed based on the next-hop node selection mechanism. Simulation results are presented to evaluate the performance of the improved GPSR protocol, which shows that the performance including packet delivery ratio and average end-to-end delay of the proposed protocol is better in some situations.

基于加权模体度的航线网络节点重要性评估指标

针对加权模体度影响航线网络节点重要性评估结果的问题,提出基于度和加权模体度的航线网络节点重要性评估指标:综合加权模体度.综合加权模体度包括综合加权三元模体度和综合加权四元模体度.引入加权聚类系数、加权度中心性、加权中介中心性和加权紧密中心性4个指标,通过网络鲁棒性评估策略和节点影响力评估策略评价综合加权模体度指标的有效性.以6家航空公司的航线网络为实验数据集,验证提出的综合加权模体度指标的有效性和适用性.研究结果表明:相比于加权聚类系数、加权度中心性、加权中介中心性、加权紧密中心性和综合加权三元模体度指标,根据综合加权四元模体度的节点重要性排序结果对6家航空公司航线网络的节点进行攻击时,网络效率和最大连通子图相对大小下降最快;以综合加权四元模体度指标计算的重要性排名前10的节点作为初始感染节点集时,6家航空公司航线网络中的免疫节点数上升速度最快;感染率在传播阈值附近取值时,综合加权四元模体度的节点重要性排序结果更接近传染病(Susceptible-Infectious-Recovered,SIR)模型的节点重要性排序结果,尤其是在网络规模和聚类系数较小的航线网络中.根据网络鲁棒性评估策略和节点影响力评估策略分析结果,综合加权四元模体度能更有效和准确地评估加权航线网络的节点重要性.

高效低碰撞的太赫兹无线定向MAC协议

针对网络中节点在争用访问时期发送消息碰撞、时隙分配机制低效和空闲时隙浪费等问题,提出了一种高效低碰撞的太赫兹无线定向MAC协议.文中提出了3种新机制,分别是基于缓存权重的分组机制、复用竞争机制以及时隙分配机制.首先,在微网控制器(Piconet Coordinator,PNC)中生成缓存映射表记录节点的缓存情况;然后,根据缓存映射表为节点分组,节点在分组对应的时隙中进行时隙请求,PNC根据节点缓存为节点分配时隙;最后,未申请到时隙的节点利用剩余时隙进行复用竞争操作.仿真结果表明,所提出的方案有效地提高了网络的吞吐量,减少了消息碰撞率并降低了平均时延.

改进AHM-RS赋权的网络节点多属性可拓聚类模型

针对网络节点模型聚类时未充分考虑网络节点的多属性以及结果片面性的问题,设计改进属性层次模型-粗糙集(Attribute hierarchy model-rough set,AHM-RS)赋权的网络节点多属性可拓聚类模型。通过改进AHMRS赋权法与可拓聚类算法,建立网络节点多属性可拓聚类模型;利用改进AHM法计算主观权重;利用RS赋权法计算客观权重;通过相对信息熵,组合主客观权重,得到综合权重;依据节点属性值和综合权重,获取节点综合重要性度量值;通过节点可拓距与综合重要性度量值,计算节点关于聚类等级的关联度,按照关联度确定该节点所属的聚类等级。实验证明:该模型可有效计算网络节点多属性综合权重,以及网络节点的关联度,完成网络节点多属性可拓聚类;在不同节点删除比例时,该模型可拓聚类的Rand指数均较高,即网络节点多属性可拓聚类精度较高。

融合用户传播风险和节点影响力分析的虚假信息传播控制方法

在线社交网络中虚假信息传播蔓延成为当前网络空间安全治理面临的重要挑战。提出一种融合用户传播风险和节点影响力分析的虚假信息传播控制方法DDC-UPRNI(disinformation diffusion control method integrating user propagation risk and node influence analysis)。综合考虑虚假信息传播特征空间的多样性和复杂性,通过自注意力机制实现用户传播虚假信息行为维度、时间维度和内容维度特征的嵌入表示,运用改进的无监督聚类K-means++算法实现不同用户传播风险等级的自动划分;设计一种自适应加权策略实现对离散粒子群优化算法的改进,进而提出一种基于离散粒子群优化的虚假信息传播关键节点选取方法,用于从具有特定传播风险等级的用户节点集合中选取若干个具有影响力的控制驱动节点,从而实现精准、高效的虚假信息传播控制;基于现实在线社交网络平台上开展试验,结果表明,所提出的DDC-UPRNI方法与现有算法相比,在控制效果和时间复杂度等重要指标上具有明显优势。该方法为社会网络空间中的虚假信息管控治理提供重要参考。

面向对象高分影像归纳式图神经网络分类法

传统面向对象分类多采用遥感影像的光谱特征,忽略了影像对象之间的空间特征。本文提出了一种采用改进归纳式图神经网络的高分遥感影像面向对象分类方法,实现了光谱-空间复合节点相似度的融合系数自适应调节,以及邻域节点采样最佳数量的自动确定。首先,改进KNN图构建方法,采用标准差信息量评价法确定用于构建光谱特征和空间特征的复合节点相似度的融合系数;然后,利用反馈曲线法确定最佳的采样邻域节点数量,使用GraphSAGE节点嵌入完成特征表达;最后,依托Softmax函数预测节点类别。以GID和BDCI2017数据集为试验数据,本文的构图方法相较于改进前的构图方法在分类精度上有所提升。本文分类方法的平均Kappa系数和总体精度分别优于CART分类树算法、GCN算法、GAT算法、LANet算法、CCTNet算法和SLCNet算法0.31、0.14、0.13、0.12、0.08、0.02和42.31%、7.4%、6.73%、8.69%、6.03%、1.52%,并且在植被和建设用地提取上具有较好的稳健性。本文方法为高分遥感影像土地覆盖分类提供了有效的工具。

基于组合加权k近邻分类的无线传感网络节点复制攻击检测方法

无线传感网络节点体积小,隐蔽性强,节点复制攻击检测的难度较大,为此提出一种基于组合加权k近邻分类的无线传感网络节点复制攻击检测方法。通过信标节点的空间位置数据与相距跳数得出各节点之间的相似程度,结合高斯径向基核函数求解未知节点的横轴、纵轴的空间坐标,确定各网络节点的空间位置;根据网络节点的属性特征与投票机制建立节点复制攻击模型,凭借组合加权k近邻分类法划分节点类型,并将结果传送至簇头节点,由簇头节点做出最后的仲裁,识别出节点复制攻击行为。仿真结果表明,所提方法的节点复制攻击检测率最大值为99.5%,最小值为97.9%,对节点复制攻击检测的耗时为5.41 s,通信开销数据包数量最大值为209个,最小值为81个。

DCE-MRI联合IVIM-DWI诊断宫颈癌淋巴结转移的价值研究

目的 分析动态增强磁共振成像(DCE-M RI)联合体素内不相干运动扩散加权成像(IVIM-DWI)对宫颈癌淋巴结转移的预测价值,为宫颈癌淋巴结转移的诊治提供参考依据。方法回顾性纳入2021年6月-2023年6月收治于新疆维吾尔自治区人民医院并经病理确诊的103例宫颈癌患者临床资料进行分析,根据是否发生盆腔淋巴结转移分为转移组(26例)和未转移组(77例)。于根治性子宫切除术前对患者进行DCE-MRI和IVIM-DWI检查,获取相关检查参数[DCE-MRI参数包括容量转运常数(K^(trans))、速率常数(K_(ep))、血管外间隙容积分数(V_(e)),IVIM-DWI参数包括表观扩散系数(ADC)、纯扩散系数(D)、伪扩散系截(D^(*))和灌注分数(f)],对比两组DCE-MRI和IVIMDWI参数的差异性,绘制受试者工作特征(ROC)曲线分析DCE-MRI联合IVIM-DWI对宫颈癌淋巴结转移的预测价值。结果在IVIM-DW参数中,转移组D值低于未转移组,f值高于未转移组(P<0.05);两组ADC、D^(*)值比较无显著性差异(P>0.05)。在DCEMRI参数中,转移组K^(trans)、V_(e)值均高于未转移组(P<0.05);两组K_(ep)值比较无显著性差异(P>0.05)。绘制ROC曲线显示,D、f、K^(trans)、V_(e)单独预测宫颈癌淋巴结转移的价值中等,曲线下面积(AUC)分别为0.842、0.749、0.785、0.828,特异度分别为0.714、0.688、0.662、0.766,敏感度分别为0.846、0.731、0.846、0.731;而各参数联合对宫颈癌淋巴结转移的预测价值较高,AUC为0.959,特异度为0.883,敏感度为0.923。结论DCE-MRI与IVIM-DWI检查获取的D、f、K^(trans)、V_(e)等参数均是预测宫颈癌淋巴结转移的理想指标,且上述参数联合对宫颈癌淋巴结转移具有更高的预测价值。

Orlicz空间中二元多项式插值的逼近

该文研究Orlicz空间中以Lissajous-Chebyshev结点为插值结点的二元多项式插值逼近问题.借助Holder不等式,光滑模等基本工具,利用Marcinkiewicz-Zygmund不等式及最佳单边逼近首先给出了Orlicz空间中高阶插值逼近的逼近度,其次研究了在不同情况下的二元插值逼近问题,利用光滑模,有界变差函数的总变差及最佳多项式逼近的阶给出了二元插值逼近度的估计.

无人机自组网快速稳定加权分簇算法

在无人机自组网中,网络规模增大会使节点间平均跳数增加,网络管理和路由协议运行更艰难。分簇结构可用来优化网络管理,提高网络的可拓展性。针对无人机高移动造成的簇结构不稳定以及分簇结构鲁棒性差的问题,提出了一种快速稳定加权分簇算法。该算法对比现有的加权分簇算法,对链路保持率、节点度和相对速度三个指标的选取进行改进。针对战场和应急场景下簇头节点掉线带来的簇振荡,提出了一种高效的簇维护机制。最后通过仿真验证该算法的性能,结果表明,与现有改进型加权分簇算法相比,该算法可以有效降低成簇的时间,同时在簇头节点掉线的情况下快速恢复,更适用于复杂环境下的网络部署。

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.

宽度学习系统中鲁棒性权值矩阵组合的筛选方法

宽度学习系统(BLS)具有出色的计算效率和预测准确性;然而,在传统BLS框架中,权值矩阵采用随机生成的方式,存在学习结果不稳定的风险。因此,设计一种BLS中鲁棒性权值矩阵组合的筛选方法(RWS-BLS)。首先,通过4组函数数据的验证,揭示随机权值矩阵在样本整体训练误差上的显著差异性;其次,研究权值矩阵组合的形式,放宽筛选条件的严格最优限制,将最优转换为较优,并将误差最小值限定在指定范围内,定义精英组合等条件;最后,得到可靠的权值矩阵的组合,有效降低随机性影响,并建立稳健的模型。实验结果表明,在16组模拟数据、NORB数据集和5组UCI回归数据集上,在数据更换或受噪声扰动的情况下,与BLS方法相比,所提方法的均方误差(MSE)下降了7.32%、8.73%和1.63%。RWS-BLS为BLS提供了一种模型平稳性研究的方向,提高了含有随机参数模型的效率和稳定性,并对涉及随机参数的其他机器学习方法具有借鉴作用。

基于多策略改进蝴蝶优化算法的无线传感网络节点覆盖优化

针对无线传感网络(WSN)的节点覆盖存在着覆盖率低、节点分布不均匀的问题,提出一种基于多策略改进的蝴蝶优化算法(MIBOA)的节点覆盖优化策略。首先,将基础的蝴蝶优化算法(BOA)与麻雀搜索算法(SSA)结合改进搜索过程;其次,引入自适应权重系数提高寻优精度和收敛速度;最后,对当前最优个体进行柯西变异扰动,提高算法鲁棒性。基准测试函数的寻优实验结果说明,MIBOA基本可在3 s内求解测试函数最优值,且收敛平均值精度较BOA提高了97.96%。将MIBOA应用于WSN节点覆盖优化问题,与BOA和SSA相比,节点覆盖率至少提高了3.63个百分点;与改进灰狼优化算法(IGWO)相比,部署时间缩短了145.82 s;与改进鲸群优化算法(IWOA)相比,节点覆盖率提高了0.20个百分点且时间缩短了1112.61 s。综上,MIBOA可较好提高节点覆盖率并降低冗余覆盖率,有效延长WSN的生存时间。