基于Graph Transformer的半监督异配图表示学习模型

现有的图卷积网络(GCN)模型基于同配性假设,无法直接应用于异配图的表示学习,且许多异配图表示学习的研究工作受消息传递机制的限制,导致节点特征混淆和特征过度挤压而出现过平滑问题。针对这些问题,提出一种基于Graph Transformer的半监督异配图表示学习模型HPGT(HeteroPhilic Graph Transformer)。首先,使用度连接概率矩阵采样节点的路径邻域,再通过自注意力机制自适应地聚合路径上的节点异配连接模式,编码得到节点的结构信息,用节点的原始属性信息和结构信息构建Transformer层的自注意力模块;其次,将每个节点自身的隐层表示与它的邻域节点的隐层表示分离更新以避免节点通过自注意力模块聚合过量的自身信息,再把每个节点表示与它的邻域表示连接,得到单个Transformer层的输出,另外,将所有的Transformer层的输出跳连到最终的节点隐层表示以防止中间层信息丢失;最后,使用线性层和Softmax层将节点的隐层表示映射到节点的预测标签。实验结果表明,与无结构编码(SE)的模型相比,基于度连接概率的SE能为Transformer层的自注意力模块提供有效的偏差信息,HPGT平均准确率提升0.99%~11.98%;与对比模型相比,在异配数据集(Texas、Cornell、Wisconsin和Actor)上,模型节点分类准确率提升0.21%~1.69%,在同配数据集(Cora、CiteSeer和PubMed)上,节点分类准确率分别达到了0.8379、0.7467和0.8862。以上结果验证了HPGT具有较强的异配图表示学习能力,尤其适用于强异配图节点分类任务。

基于RoBERTa和图增强Transformer的序列推荐方法

自推荐系统出现以来,有限的数据信息就一直制约着推荐算法的进一步发展。为降低数据稀疏性的影响,增强非评分数据的利用率,基于神经网络的文本推荐模型相继被提出,但主流的卷积或循环神经网络在文本语义理解和长距离关系捕捉方面存在明显劣势。为了更好地挖掘用户与商品之间的深层潜在特征,进一步提高推荐质量,提出一种基于Ro BERTa和图增强Transformer的序列推荐(RGT)模型。引入评论文本数据,首先利用预训练的Ro BERTa模型捕获评论文本中的字词语义特征,初步建模用户的个性化兴趣,然后根据用户与商品的历史交互信息,构建具有时序特性的商品关联图注意力机制网络模型,通过图增强Transformer的方法将图模型学习到的各个商品的特征表示以序列的形式输入Transformer编码层,最后将得到的输出向量与之前捕获的语义表征以及计算得到的商品关联图的全图表征输入全连接层,以捕获用户全局的兴趣偏好,实现用户对商品的预测评分。在3组真实亚马逊公开数据集上的实验结果表明,与Deep FM、Conv MF等经典文本推荐模型相比,RGT模型在均方根误差(RMSE)和平均绝对误差(MAE)2种指标上有显著提升,相较于最优对比模型最高分别提升4.7%和5.3%。

Brain Functional Network Generation Using Distribution-Regularized Adversarial Graph Autoencoder with Transformer for Dementia Diagnosis

The topological connectivity information derived from the brain functional network can bring new insights for diagnosing and analyzing dementia disorders.The brain functional network is suitable to bridge the correlation between abnormal connectivities and dementia disorders.However,it is challenging to access considerable amounts of brain functional network data,which hinders the widespread application of data-driven models in dementia diagnosis.In this study,a novel distribution-regularized adversarial graph auto-Encoder(DAGAE)with transformer is proposed to generate new fake brain functional networks to augment the brain functional network dataset,improving the dementia diagnosis accuracy of data-driven models.Specifically,the label distribution is estimated to regularize the latent space learned by the graph encoder,which canmake the learning process stable and the learned representation robust.Also,the transformer generator is devised to map the node representations into node-to-node connections by exploring the long-term dependence of highly-correlated distant brain regions.The typical topological properties and discriminative features can be preserved entirely.Furthermore,the generated brain functional networks improve the prediction performance using different classifiers,which can be applied to analyze other cognitive diseases.Attempts on the Alzheimer’s Disease Neuroimaging Initiative(ADNI)dataset demonstrate that the proposed model can generate good brain functional networks.The classification results show adding generated data can achieve the best accuracy value of 85.33%,sensitivity value of 84.00%,specificity value of 86.67%.The proposed model also achieves superior performance compared with other related augmentedmodels.Overall,the proposedmodel effectively improves cognitive disease diagnosis by generating diverse brain functional networks.

Micro-expression recognition algorithm based on graph convolutional network and Transformer model

Micro-expressions are spontaneous, unconscious movements that reveal true emotions.Accurate facial movement information and network training learning methods are crucial for micro-expression recognition.However, most existing micro-expression recognition technologies so far focus on modeling the single category of micro-expression images and neural network structure.Aiming at the problems of low recognition rate and weak model generalization ability in micro-expression recognition, a micro-expression recognition algorithm is proposed based on graph convolution network(GCN) and Transformer model.Firstly, action unit(AU) feature detection is extracted and facial muscle nodes in the neighborhood are divided into three subsets for recognition.Then, graph convolution layer is used to find the layout of dependencies between AU nodes of micro-expression classification.Finally, multiple attentional features of each facial action are enriched with Transformer model to include more sequence information before calculating the overall correlation of each region.The proposed method is validated in CASME II and CAS(ME)^2 datasets, and the recognition rate reached 69.85%.

伪时空图卷积网络修复姿态引导的Transformer行人视频修复方法

为解决监控视频中被遮挡行人的修复问题,提出了一种基于人体姿态的行人视频修复方法,即先修复视频中残缺的行人姿态序列,然后在修补后的姿势序列的引导下修复视频帧中人体的缺失部分.该方法采用OpenPose从视频中提取被遮挡的人体姿态序列,针对其因存在遮挡情况导致未识别出和未准确识别部分关节点的问题,提出了一种伪时空图卷积网络模型对缺失姿态进行修复,得到一个相对准确的姿态序列;基于修复后的姿态,提出了基于姿态序列引导的Transformer行人视频修复模型.在Human3.6M数据集上进行了测试,所提出的方法在4个指标PSNR,RMSE,SSIM,LPIPS上均比对比方法有提升,特别是RMSE指标提升了9.50%,LPIPS指标提升了21.67%.

Network Configuration Entity Extraction Method Based on Transformer with Multi-Head Attention Mechanism

Nowadays,ensuring thequality of networkserviceshas become increasingly vital.Experts are turning toknowledge graph technology,with a significant emphasis on entity extraction in the identification of device configurations.This research paper presents a novel entity extraction method that leverages a combination of active learning and attention mechanisms.Initially,an improved active learning approach is employed to select the most valuable unlabeled samples,which are subsequently submitted for expert labeling.This approach successfully addresses the problems of isolated points and sample redundancy within the network configuration sample set.Then the labeled samples are utilized to train the model for network configuration entity extraction.Furthermore,the multi-head self-attention of the transformer model is enhanced by introducing the Adaptive Weighting method based on the Laplace mixture distribution.This enhancement enables the transformer model to dynamically adapt its focus to words in various positions,displaying exceptional adaptability to abnormal data and further elevating the accuracy of the proposed model.Through comparisons with Random Sampling(RANDOM),Maximum Normalized Log-Probability(MNLP),Least Confidence(LC),Token Entrop(TE),and Entropy Query by Bagging(EQB),the proposed method,Entropy Query by Bagging and Maximum Influence Active Learning(EQBMIAL),achieves comparable performance with only 40% of the samples on both datasets,while other algorithms require 50% of the samples.Furthermore,the entity extraction algorithm with the Adaptive Weighted Multi-head Attention mechanism(AW-MHA)is compared with BILSTM-CRF,Mutil_Attention-Bilstm-Crf,Deep_Neural_Model_NER and BERT_Transformer,achieving precision rates of 75.98% and 98.32% on the two datasets,respectively.Statistical tests demonstrate the statistical significance and effectiveness of the proposed algorithms in this paper.

SGT:Session-based Recommendation with GRU and Transformer

Session-based recommendation aims to predict user preferences based on anonymous behavior sequences.Recent research on session-based recommendation systems has mainly focused on utilizing attention mechanisms on sequential patterns,which has achieved significant results.However,most existing studies only consider individual items in a session and do not extract information from continuous items,which can easily lead to the loss of information on item transition relationships.Therefore,this paper proposes a session-based recommendation algorithm(SGT)based on Gated Recurrent Unit(GRU)and Transformer,which captures user interests by learning continuous items in the current session and utilizes all item transitions on sessions in a more refined way.By combining short-term sessions and long-term behavior,user dynamic preferences are captured.Extensive experiments were conducted on three session-based recommendation datasets,and compared to the baseline methods,both the recall rate Recall@20 and the mean reciprocal rank MRR@20 of the SGT algorithm were improved,demonstrating the effectiveness of the SGT method.

基于Swin Transformer和图形推理的结直肠息肉分割方法

针对结直肠息肉图像分割中病灶区域尺度变化大、边缘模糊以及息肉与正常组织对比度低等问题,导致病变区域分割精度低和分割边界存在伪影,提出一种基于Swin Transformer和图形推理的自适应网络.该网络一是利用Swin Transformer编码器逐层提取输入图像的全局上下文信息,弱化背景信息干扰,多尺度分析病变区域的显著性特点.二是提出全局与局部特征交互模块增强网络对复杂病灶的空间感知能力,突出待分割目标的关键位置信息,精准定位目标.三是通过区域引导图推理模块以图循环递推的方式挖掘先验信息之间的高阶显性关系,促进图间信息传递.四是设计面向边缘细节的边缘约束图推理模块,整合边缘细节,改善分割效果,提高分割精度.在CVC-ClinicDB、Kvasir、CVC-ColonDB和ETIS数据集上进行实验,其Dice系数分别为0.939,0.926,0.810和0.788,平均交并比分别为0.889,0.879,0.731和0.710,分割性能优于现有方法.仿真实验结果表明,对于形态结构复杂、对比度低和边缘模糊的结直肠息肉图像均有较高的分割精度.

融合Transformer和语义图卷积的三维人体姿态估计方法

为了进一步提升从单目二维人体姿态预测三维人体姿态的方法性能,提出一种融合Transformer和语义图卷积的三维人体姿态估计模型,模型由4个部分组成,Transformer编码网络、语义图卷积编码网络、姿态坐标预测模块和姿态坐标错误回归模块。首先,Transformer编码网络对关节特征进行全局特征编码,以增强人体姿态的全局关联性。其次,语义图卷积编码网络专注于局部关节特征提取,以加强局部关节特征之间的关联性。接下来,姿态坐标预测模块和姿态坐标错误回归模块将关节全局和局部编码特征融合,以增强对三维姿态的准确建模能力。通过在Human3.6M数据集上进行实验表明,方法在估计性能方面取得了较好的改进,以真实的二维人体姿态作为输入,在MPJPE和PA-MPJPE值分别为32.7和25.9 mm,与实验对照方法相比,性能分别提升了3.82%和1.14%。

基于图Transformer网络的城市路网短时交通流预测模型

针对城市路网短时交通流预测问题,在考虑路网交通状态时空相关性基础上,提出了一种基于图Transformer(graph transformer,Graformer)的预测方法。该方法将多条路段的交通状态预测问题转化为图节点状态预测问题,针对区分相同结构的空间路网结构图,将带有边的图同构网络(graph isomorphism network with edges,GINE)和Transformer网络相结合,对交通状态在路网层面的时空相关性进行建模,从而实现城市路网短时交通流预测。具体来说,Graformer模型首先利用长短期记忆网络(long short-term memory,LSTM)对交通数据的时序信息进行预处理,接着采用基于GINE与Transformer的全局注意力机制提取交通数据的空间特征,最后实现路网各路段交通流的同步预测。

基于骨架动作识别的协作卷积Transformer网络

近年来,基于骨架的人体动作识别任务因骨架数据的鲁棒性和泛化能力而受到了广泛关注。其中,将人体骨骼建模为时空图的图卷积网络取得了显著的性能。然而图卷积主要通过一系列3D卷积来学习长期交互联系,这种联系偏向于局部并且受到卷积核大小的限制,无法有效地捕获远程依赖关系。该文提出一种协作卷积Transformer网络(Co-ConvT),通过引入Transformer中的自注意力机制建立远程依赖关系,并将其与图卷积神经网络(GCNs)相结合进行动作识别,使模型既能通过图卷积神经网络提取局部信息,也能通过Transformer捕获丰富的远程依赖项。另外,Transformer的自注意力机制在像素级进行计算,因此产生了极大的计算代价,该模型通过将整个网络分为两个阶段,第1阶段使用纯卷积来提取浅层空间特征,第2阶段使用所提出的ConvT块捕获高层语义信息,降低了计算复杂度。此外,原始Transformer中的线性嵌入被替换为卷积嵌入,获得局部空间信息增强,并由此去除了原始模型中的位置编码,使模型更轻量。在两个大规模权威数据集NTU-RGB+D和Kinetics-Skeleton上进行实验验证,该模型分别达到了88.1%和36.6%的Top-1精度。实验结果表明,该模型的性能有了很大的提高。

融合自适应图卷积与Transformer序列模型的中文手语翻译方法

针对手语翻译方法所存在的动作特征提取以及时序翻译方面存在的问题,提出一种融合自适应图卷积AGCN与Transformer时序模型的AGCN-T手语翻译网络。自适应图卷积网络用于学习手语动作中骨骼节点的交互空间依赖信息;Transformer时序模块捕捉手语动作序列的时间关系特征信息并将其翻译成可理解的手语内容。此外,在预处理部分,提出了一种移动窗口的关键帧提取算法,并用MediaPipe姿态估计算法对关键帧图像序列进行骨架提取。实验表明,该方法在大型中文连续手语数据集CCSL的词错率达到了3.75%,精度为97.87%,优于其他先进的手语翻译方法。

基于集成GCN-Transformer网络的ENSO预测模型

厄尔尼诺-南方涛动(El Niño-Southern Oscillation, ENSO)是热带太平洋海表面温度发生异常的现象,会导致冰雹、洪水、台风等极端天气的出现,因此对ENSO的准确预测意义重大。本文设计了集成GCN-Transformer(GCNTR)模型,首先运用Transformer网络的全局特征聚焦能力对数据特征进行编码,然后结合GCN提取图数据特征的能力,最后引入特征融合门控机制将经过编码的特征和GCN提取的特征进行融合,实现ENSO的精确预测。结果表明,GCNTR模型实现了对ENSO提前20个月的预测,比ENSOTR多了3个月,比Transformer多了5个月,并且模型绝大部分的预测精度优于其他模型。与现有的方法相比,GCNTR模型能够实现对ENSO更好的预测。

Graph Transformer for Communities Detection in Social Networks

Graphs are used in various disciplines such as telecommunication,biological networks,as well as social networks.In large-scale networks,it is challenging to detect the communities by learning the distinct properties of the graph.As deep learning hasmade contributions in a variety of domains,we try to use deep learning techniques to mine the knowledge from large-scale graph networks.In this paper,we aim to provide a strategy for detecting communities using deep autoencoders and obtain generic neural attention to graphs.The advantages of neural attention are widely seen in the field of NLP and computer vision,which has low computational complexity for large-scale graphs.The contributions of the paper are summarized as follows.Firstly,a transformer is utilized to downsample the first-order proximities of the graph into a latent space,which can result in the structural properties and eventually assist in detecting the communities.Secondly,the fine-tuning task is conducted by tuning variant hyperparameters cautiously,which is applied to multiple social networks(Facebook and Twitch).Furthermore,the objective function(crossentropy)is tuned by L0 regularization.Lastly,the reconstructed model forms communities that present the relationship between the groups.The proposed robust model provides good generalization and is applicable to obtaining not only the community structures in social networks but also the node classification.The proposed graph-transformer shows advanced performance on the social networks with the average NMIs of 0.67±0.04,0.198±0.02,0.228±0.02,and 0.68±0.03 on Wikipedia crocodiles,Github Developers,Twitch England,and Facebook Page-Page networks,respectively.

A Model Transformation Approach for Detecting Distancing Violations in Weighted Graphs

This work presents the design of an Internet of Things(IoT)edge-based system based on model transformation and complete weighted graph to detect violations of social distancing measures in indoor public places.Awireless sensor network based on Bluetooth Low Energy is introduced as the infrastructure of the proposed design.A hybrid model transformation strategy for generating a graph database to represent groups of people is presented as a core middleware layer of the detecting system’s proposed architectural design.A Neo4j graph database is used as a target implementation generated from the proposed transformational system to store all captured real-time IoT data about the distances between individuals in an indoor area and answer user predefined queries,expressed using Neo4j Cypher,to provide insights from the stored data for decision support.As proof of concept,a discrete-time simulation model was adopted for the design of a COVID-19 physical distancing measures case study to evaluate the introduced system architecture.Twenty-one weighted graphs were generated randomly and the degrees of violation of distancing measures were inspected.The experimental results demonstrate the capability of the proposed system design to detect violations of COVID-19 physical distancing measures within an enclosed area.

Zeta Functions of the Complement and xyz-Transformations of a Regular Graph

Let Z(λ,G)denote the zeta function of a graph G.In this paper the complement G^Cand the G^(xyz)-transformation G^(xyz)of an r-regular graph G with n vertices and m edges for x,y,z∈{0,1,+,-},are considerd.The relationship between Z(λ,G)and Z(λ,G^C)is obtained.For all x,y,z∈{0,1,+,-},the explicit formulas for the reciprocal of Z(λ,G^(xyz))in terms of r,m,n and the characteristic polynomial of G are obtained.Due to limited space,only the expressions for G^(xyz)with z=0,and xyz∈{0++,+++,1+-}are presented here.

Markov Chains Based on Random Generalized 1-Flipper Operations for Connected Regular Multi-digraphs

The properties of generalized flip Markov chains on connected regular digraphs are discussed.The 1-Flipper operation on Markov chains for undirected graphs is generalized to that for multi-digraphs.The generalized 1-Flipper operation preserves the regularity and weak connectivity of multi-digraphs.The generalized 1-Flipper operation is proved to be symmetric.Moreover,it is presented that a series of random generalized 1-Flipper operations eventually lead to a uniform probability distribution over all connected d-regular multi-digraphs without loops.

End-to-end aspect category sentiment analysis based on type graph convolutional networks

For the existing aspect category sentiment analysis research,most of the aspects are given for sentiment extraction,and this pipeline method is prone to error accumulation,and the use of graph convolutional neural network for aspect category sentiment analysis does not fully utilize the dependency type information between words,so it cannot enhance feature extraction.This paper proposes an end-to-end aspect category sentiment analysis(ETESA)model based on type graph convolutional networks.The model uses the bidirectional encoder representation from transformers(BERT)pretraining model to obtain aspect categories and word vectors containing contextual dynamic semantic information,which can solve the problem of polysemy;when using graph convolutional network(GCN)for feature extraction,the fusion operation of word vectors and initialization tensor of dependency types can obtain the importance values of different dependency types and enhance the text feature representation;by transforming aspect category and sentiment pair extraction into multiple single-label classification problems,aspect category and sentiment can be extracted simultaneously in an end-to-end way and solve the problem of error accumulation.Experiments are tested on three public datasets,and the results show that the ETESA model can achieve higher Precision,Recall and F1 value,proving the effectiveness of the model.

基于多尺度图Transformer的滚动轴承故障诊断

针对传统图神经网络在故障诊断中使用单一尺度进行特征提取且难以在复杂工况下提取信号的弱特征问题,提出了一种基于多尺度图Transformer的滚动轴承故障诊断方法。该方法提出了一种新的图节点多尺度特征聚合模块,扩大特征提取的感受野以增强特征表示;构建了图节点的中心性编码和空间性编码,以获得图结构信息;利用多头自注意力对故障节点进行特征提取和学习,提高方法捕捉重要特征的能力。在凯斯西储大学轴承数据集和滚动轴承实验平台上分别进行实验验证,诊断准确率最高为99.86%,平均准确率也在98%以上。结果表明,提出的多尺度图Transformer网络模型在多种工况下均能准确的进行故障分类。

国内外教育数字化转型研究的比较分析

教育数字化转型是通过综合应用数字技术,掌握数字化能力,构建智慧教育生态和数字治理体系的过程。为比较国内外教育数字化转型领域的研究成果和特征,以Web of Science核心合集和CNKI数据库为数据源,借助CiteSpace可视化软件,围绕文献作者、机构、关键词等信息进行聚类分析。研究发现:在教育数字化转型领域,国内发文量自2022年迅速上升,发文最多的机构是华东师范大学开放教育学院。国外发文量最高的国家是西班牙,最多的机构是墨西哥蒙特雷技术与高等教育学院。国外的研究热点主要侧重高等教育体系,呈现相对稳定的演进趋势;国内更关注职业教育,并逐步衍生出部分新的研究主题。国内存在主要问题包括基础设施建设急需标准规范体系引领,数字化技术与教育教学缺乏深度融合,教育评估与监测缺少成熟度评估规范。