The fluctuation of wind power affects the operating safety and power consumption of the electric power grid and restricts the grid connection of wind power on a large scale.Therefore,wind power forecasting plays a key...The fluctuation of wind power affects the operating safety and power consumption of the electric power grid and restricts the grid connection of wind power on a large scale.Therefore,wind power forecasting plays a key role in improving the safety and economic benefits of the power grid.This paper proposes a wind power predicting method based on a convolutional graph attention deep neural network with multi-wind farm data.Based on the graph attention network and attention mechanism,the method extracts spatial-temporal characteristics from the data of multiple wind farms.Then,combined with a deep neural network,a convolutional graph attention deep neural network model is constructed.Finally,the model is trained with the quantile regression loss function to achieve the wind power deterministic and probabilistic prediction based on multi-wind farm spatial-temporal data.A wind power dataset in the U.S.is taken as an example to demonstrate the efficacy of the proposed model.Compared with the selected baseline methods,the proposed model achieves the best prediction performance.The point prediction errors(i.e.,root mean square error(RMSE)and normalized mean absolute percentage error(NMAPE))are 0.304 MW and 1.177%,respectively.And the comprehensive performance of probabilistic prediction(i.e.,con-tinuously ranked probability score(CRPS))is 0.580.Thus,the significance of multi-wind farm data and spatial-temporal feature extraction module is self-evident.展开更多
Continuous sign language recognition(CSLR)is challenging due to the complexity of video background,hand gesture variability,and temporal modeling difficulties.This work proposes a CSLR method based on a spatialtempora...Continuous sign language recognition(CSLR)is challenging due to the complexity of video background,hand gesture variability,and temporal modeling difficulties.This work proposes a CSLR method based on a spatialtemporal graph attention network to focus on essential features of video series.The method considers local details of sign language movements by taking the information on joints and bones as inputs and constructing a spatialtemporal graph to reflect inter-frame relevance and physical connections between nodes.The graph-based multihead attention mechanism is utilized with adjacent matrix calculation for better local-feature exploration,and short-term motion correlation modeling is completed via a temporal convolutional network.We adopted BLSTM to learn the long-termdependence and connectionist temporal classification to align the word-level sequences.The proposed method achieves competitive results regarding word error rates(1.59%)on the Chinese Sign Language dataset and the mean Jaccard Index(65.78%)on the ChaLearn LAP Continuous Gesture Dataset.展开更多
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 signi...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.展开更多
Automatic text summarization(ATS)plays a significant role in Natural Language Processing(NLP).Abstractive summarization produces summaries by identifying and compressing the most important information in a document.Ho...Automatic text summarization(ATS)plays a significant role in Natural Language Processing(NLP).Abstractive summarization produces summaries by identifying and compressing the most important information in a document.However,there are only relatively several comprehensively evaluated abstractive summarization models that work well for specific types of reports due to their unstructured and oral language text characteristics.In particular,Chinese complaint reports,generated by urban complainers and collected by government employees,describe existing resident problems in daily life.Meanwhile,the reflected problems are required to respond speedily.Therefore,automatic summarization tasks for these reports have been developed.However,similar to traditional summarization models,the generated summaries still exist problems of informativeness and conciseness.To address these issues and generate suitably informative and less redundant summaries,a topic-based abstractive summarization method is proposed to obtain global and local features.Additionally,a heterogeneous graph of the original document is constructed using word-level and topic-level features.Experiments and analyses on public review datasets(Yelp and Amazon)and our constructed dataset(Chinese complaint reports)show that the proposed framework effectively improves the performance of the abstractive summarization model for Chinese complaint reports.展开更多
With the continuous improvement of the e-commerce ecosystem and the rapid growth of e-commerce data, inthe context of the e-commerce ecosystem, consumers ask hundreds of millions of questions every day. In order toimp...With the continuous improvement of the e-commerce ecosystem and the rapid growth of e-commerce data, inthe context of the e-commerce ecosystem, consumers ask hundreds of millions of questions every day. In order toimprove the timeliness of customer service responses, many systems have begun to use customer service robotsto respond to consumer questions, but the current customer service robots tend to respond to specific questions.For many questions that lack background knowledge, they can generate only responses that are biased towardsgenerality and repetitiveness. To better promote the understanding of dialogue and generate more meaningfulresponses, this paper introduces knowledge information into the research of question answering system by usinga knowledge graph. The unique structured knowledge base of the knowledge graph is convenient for knowledgequery, can acquire knowledge faster, and improves the background information needed for answering questions. Toavoid the lack of information in the dialogue process, this paper proposes the Multi-hop Knowledge InformationEnhanced Dialogue-Graph Attention (MKIED-GA) model. The model first retrieves the problem subgraph directlyrelated to the input information from the entire knowledge base and then uses the graph neural network as theknowledge inference module on the subgraph to encode the subgraph. The graph attention mechanism is usedto determine the one-hop and two-hop entities that are more relevant to the problem to achieve the aggregationof highly relevant neighbor information. This further enriches the semantic information to provide a betterunderstanding of the meaning of the input question and generate appropriate response information. In the processof generating a response, a multi-attention flow mechanism is used to focus on different information to promotethe generation of better responses. Experiments have proved that the model presented in this article can generatemore meaningful responses than other models.展开更多
PM2.5 concentration prediction is of great significance to environmental protection and human health.Achieving accurate prediction of PM2.5 concentration has become an important research task.However,PM2.5 pollutants ...PM2.5 concentration prediction is of great significance to environmental protection and human health.Achieving accurate prediction of PM2.5 concentration has become an important research task.However,PM2.5 pollutants can spread in the earth’s atmosphere,causing mutual influence between different cities.To effectively capture the air pollution relationship between cities,this paper proposes a novel spatiotemporal model combining graph attention neural network(GAT)and gated recurrent unit(GRU),named GAT-GRU for PM2.5 concentration prediction.Specifically,GAT is used to learn the spatial dependence of PM2.5 concentration data in different cities,and GRU is to extract the temporal dependence of the long-term data series.The proposed model integrates the learned spatio-temporal dependencies to capture long-term complex spatio-temporal features.Considering that air pollution is related to the meteorological conditions of the city,the knowledge acquired from meteorological data is used in the model to enhance PM2.5 prediction performance.The input of the GAT-GRU model consists of PM2.5 concentration data and meteorological data.In order to verify the effectiveness of the proposed GAT-GRU prediction model,this paper designs experiments on real-world datasets compared with other baselines.Experimental results prove that our model achieves excellent performance in PM2.5 concentration prediction.展开更多
Event detection(ED)seeks to recognize event triggers and classify them into the predefined event types.Chinese ED is formulated as a character-level task owing to the uncertain word boundaries.Prior methods try to inc...Event detection(ED)seeks to recognize event triggers and classify them into the predefined event types.Chinese ED is formulated as a character-level task owing to the uncertain word boundaries.Prior methods try to incorpo-rate word-level information into characters to enhance their semantics.However,they experience two problems.First,they fail to incorporate word-level information into each character the word encompasses,causing the insufficient word-charac-ter interaction problem.Second,they struggle to distinguish events of similar types with limited annotated instances,which is called the event confusing problem.This paper proposes a novel model named Label-Aware Heterogeneous Graph Attention Network(L-HGAT)to address these two problems.Specifically,we first build a heterogeneous graph of two node types and three edge types to maximally preserve word-character interactions,and then deploy a heterogeneous graph attention network to enhance the semantic propagation between characters and words.Furthermore,we design a pushing-away game to enlarge the predicting gap between the ground-truth event type and its confusing counterpart for each character.Experimental results show that our L-HGAT model consistently achieves superior performance over prior competitive methods.展开更多
Due to coexistence of huge number of structural isomers,global search for the ground-state structures of atomic clusters is a challenging issue.The difficulty also originates from the computational cost of ab initio m...Due to coexistence of huge number of structural isomers,global search for the ground-state structures of atomic clusters is a challenging issue.The difficulty also originates from the computational cost of ab initio methods for describing the potential energy surface.Recently,machine learning techniques have been widely utilized to accelerate materials discovery and molecular simulation.Compared to the commonly used artificial neural network,graph network is naturally suitable for clusters with flexible geometric environment of each atom.Herein we develop a cluster graph attention network(CGANet)by aggregating information of neighboring vertices and edges using attention mechanism,which can precisely predict the binding energy and force of silver clusters with root mean square error of 5.4 meV/atom and mean absolute error of 42.3 meV/Å,respectively.As a proof-of-concept,we have performed global optimization of mediumsized Agn clusters(n=14–26)by combining CGANet and genetic algorithm.The reported ground-state structures for n=14–21,have been successfully reproduced,while entirely new lowest-energy structures are obtained for n=22–26.In addition to the description of potential energy surface,the CGANet is also applied to predict the electronic properties of clusters,such as HOMO energy and HOMO-LUMO gap.With accuracy comparable to ab initio methods and acceleration by at least two orders of magnitude,CGANet holds great promise in global search of lowest-energy structures of large clusters and inverse design of functional clusters.展开更多
Recent advances in spatially resolved transcriptomic technologies have enabled unprecedented opportunities to elucidate tissue architecture and function in situ.Spatial transcriptomics can provide multimodal and compl...Recent advances in spatially resolved transcriptomic technologies have enabled unprecedented opportunities to elucidate tissue architecture and function in situ.Spatial transcriptomics can provide multimodal and complementary information simultaneously,including gene expression profiles,spatial locations,and histology images.However,most existing methods have limitations in efficiently utilizing spatial information and matched high-resolution histology images.To fully leverage the multi-modal information,we propose a SPAtially embedded Deep Attentional graph Clustering(SpaDAC)method to identify spatial domains while reconstructing denoised gene expression profiles.This method can efficiently learn the low-dimensional embeddings for spatial transcriptomics data by constructing multi-view graph modules to capture both spatial location connectives and morphological connectives.Benchmark results demonstrate that SpaDAC outperforms other algorithms on several recent spatial transcriptomics datasets.SpaDAC is a valuable tool for spatial domain detection,facilitating the comprehension of tissue architecture and cellular microenvironment.The source code of SpaDAC is freely available at Github(https://github.com/huoyuying/SpaDAC.git).展开更多
Social robot accounts controlled by artificial intelligence or humans are active in social networks,bringing negative impacts to network security and social life.Existing social robot detection methods based on graph ...Social robot accounts controlled by artificial intelligence or humans are active in social networks,bringing negative impacts to network security and social life.Existing social robot detection methods based on graph neural networks suffer from the problem of many social network nodes and complex relationships,which makes it difficult to accurately describe the difference between the topological relations of nodes,resulting in low detection accuracy of social robots.This paper proposes a social robot detection method with the use of an improved neural network.First,social relationship subgraphs are constructed by leveraging the user’s social network to disentangle intricate social relationships effectively.Then,a linear modulated graph attention residual network model is devised to extract the node and network topology features of the social relation subgraph,thereby generating comprehensive social relation subgraph features,and the feature-wise linear modulation module of the model can better learn the differences between the nodes.Next,user text content and behavioral gene sequences are extracted to construct social behavioral features combined with the social relationship subgraph features.Finally,social robots can be more accurately identified by combining user behavioral and relationship features.By carrying out experimental studies based on the publicly available datasets TwiBot-20 and Cresci-15,the suggested method’s detection accuracies can achieve 86.73%and 97.86%,respectively.Compared with the existing mainstream approaches,the accuracy of the proposed method is 2.2%and 1.35%higher on the two datasets.The results show that the method proposed in this paper can effectively detect social robots and maintain a healthy ecological environment of social networks.展开更多
Essential proteins play an important role in disease diagnosis and drug development.Many methods have been devoted to the essential protein prediction by using some kinds of biological information.However,they either ...Essential proteins play an important role in disease diagnosis and drug development.Many methods have been devoted to the essential protein prediction by using some kinds of biological information.However,they either ignore the noise presented in the biological information itself or the noise generated during feature extraction.To overcome these problems,in this paper,we propose a novel method for predicting essential proteins called attention gate-graph attention network and temporal convolutional network(AG-GATCN).In AG-GATCN method,we use improved temporal convolutional network(TCN)to extract features from gene expression sequence.To address the noise in the gene expression sequence itself and the noise generated after the dilated causal convolution,we introduce attention mechanism and gating mechanism in TCN.In addition,we use graph attention network(GAT)to extract protein–protein interaction(PPI)network features,in which we construct the feature matrix by introducing node2vec technique and 7 centrality metrics,and to solve the GAT oversmoothing problem,we introduce gated tanh unit(GTU)in GAT.Finally,two types of features are integrated by us to predict essential proteins.Compared with the existing methods for predicting essential proteins,the experimental results show that AG-GATCN achieves better performance.展开更多
Sensors produce a large amount of multivariate time series data to record the states of Internet of Things(IoT)systems.Multivariate time series timestamp anomaly detection(TSAD)can identify timestamps of attacks and m...Sensors produce a large amount of multivariate time series data to record the states of Internet of Things(IoT)systems.Multivariate time series timestamp anomaly detection(TSAD)can identify timestamps of attacks and malfunctions.However,it is necessary to determine which sensor or indicator is abnormal to facilitate a more detailed diagnosis,a process referred to as fine-grained anomaly detection(FGAD).Although further FGAD can be extended based on TSAD methods,existing works do not provide a quantitative evaluation,and the performance is unknown.Therefore,to tackle the FGAD problem,this paper first verifies that the TSAD methods achieve low performance when applied to the FGAD task directly because of the excessive fusion of features and the ignoring of the relationship’s dynamic changes between indicators.Accordingly,this paper proposes a mul-tivariate time series fine-grained anomaly detection(MFGAD)framework.To avoid excessive fusion of features,MFGAD constructs two sub-models to independently identify the abnormal timestamp and abnormal indicator instead of a single model and then combines the two kinds of abnormal results to detect the fine-grained anomaly.Based on this framework,an algorithm based on Graph Attention Neural Network(GAT)and Attention Convolutional Long-Short Term Memory(A-ConvLSTM)is proposed,in which GAT learns temporal features of multiple indicators to detect abnormal timestamps and A-ConvLSTM captures the dynamic relationship between indicators to identify abnormal indicators.Extensive simulations on a real-world dataset demonstrate that the proposed algorithm can achieve a higher F1 score and hit rate than the extension of existing TSAD methods with the benefit of two independent sub-models for timestamp and indicator detection.展开更多
Due to the time-varying topology and possible disturbances in a conflict environment,it is still challenging to maintain the mission performance of flying Ad hoc networks(FANET),which limits the application of Unmanne...Due to the time-varying topology and possible disturbances in a conflict environment,it is still challenging to maintain the mission performance of flying Ad hoc networks(FANET),which limits the application of Unmanned Aerial Vehicle(UAV)swarms in harsh environments.This paper proposes an intelligent framework to quickly recover the cooperative coveragemission by aggregating the historical spatio-temporal network with the attention mechanism.The mission resilience metric is introduced in conjunction with connectivity and coverage status information to simplify the optimization model.A spatio-temporal node pooling method is proposed to ensure all node location features can be updated after destruction by capturing the temporal network structure.Combined with the corresponding Laplacian matrix as the hyperparameter,a recovery algorithm based on the multi-head attention graph network is designed to achieve rapid recovery.Simulation results showed that the proposed framework can facilitate rapid recovery of the connectivity and coverage more effectively compared to the existing studies.The results demonstrate that the average connectivity and coverage results is improved by 17.92%and 16.96%,respectively compared with the state-of-the-art model.Furthermore,by the ablation study,the contributions of each different improvement are compared.The proposed model can be used to support resilient network design for real-time mission execution.展开更多
Event temporal relation extraction is an important part of natural language processing.Many models are being used in this task with the development of deep learning.However,most of the existing methods cannot accurate...Event temporal relation extraction is an important part of natural language processing.Many models are being used in this task with the development of deep learning.However,most of the existing methods cannot accurately obtain the degree of association between different tokens and events,and event-related information cannot be effectively integrated.In this paper,we propose an event information integration model that integrates event information through multilayer bidirectional long short-term memory(Bi-LSTM)and attention mechanism.Although the above scheme can improve the extraction performance,it can still be further optimized.To further improve the performance of the previous scheme,we propose a novel relational graph attention network that incorporates edge attributes.In this approach,we first build a semantic dependency graph through dependency parsing,model a semantic graph that considers the edges’attributes by using top-k attention mechanisms to learn hidden semantic contextual representations,and finally predict event temporal relations.We evaluate proposed models on the TimeBank-Dense dataset.Compared to previous baselines,the Micro-F1 scores obtained by our models improve by 3.9%and 14.5%,respectively.展开更多
For a long time, the detection and extraction of printed surfacedefects has been a hot issue in the print industry. Nowadays, defect detectionof a large number of products still relies on traditional image processinga...For a long time, the detection and extraction of printed surfacedefects has been a hot issue in the print industry. Nowadays, defect detectionof a large number of products still relies on traditional image processingalgorithms such as scale invariant feature transform (SIFT) and orientedfast and rotated brief (ORB), and researchers need to design algorithms forspecific products. At present, a large number of defect detection algorithmsbased on object detection have been applied but need lots of labeling sampleswith defects. Besides, there are many kinds of defects in printed surface,so it is difficult to enumerate all defects. Most defect detection based onunsupervised learning of positive samples use generative adversarial networks(GAN) and variational auto-encoders (VAE) algorithms, but these methodsare not effective for complex printed surface. Aiming at these problems, Inthis paper, an unsupervised defect detection and extraction algorithm forprinted surface based on positive samples in the complex printed surface isproposed innovatively. We propose a kind of defect detection and extractionnetwork based on image matching network. This network is divided into thefull convolution network of feature points extraction, and the graph attentionnetwork using self attention and cross attention. Though the key pointsextraction network, we can get robustness key points in the complex printedimages, and the graph network can solve the problem of the deviation becauseof different camera positions and the influence of defect in the differentproduction lines. Just one positive sample image is needed as the benchmarkto detect the defects. The algorithm in this paper has been proved in “TheFirst ZhengTu Cup on Campus Machine Vision AI Competition” and gotexcellent results in the finals. We are working with the company to apply it inproduction.展开更多
We present GeoGlue,a novel method using high-resolution UAV imagery for accurate feature matching,which is normally challenging due to the complicated scenes.Current feature detection methods are performed without gui...We present GeoGlue,a novel method using high-resolution UAV imagery for accurate feature matching,which is normally challenging due to the complicated scenes.Current feature detection methods are performed without guidance of geometric priors(e.g.,geometric lines),lacking enough attention given to salient geometric features which are indispensable for accurate matching due to their stable existence across views.In this work,geometric lines arefirstly detected by a CNN-based geometry detector(GD)which is pre-trained in a self-supervised manner through automatically generated images.Then,geometric lines are naturally vectorized based on GD and thus non-significant features can be disregarded as judged by their disordered geometric morphology.A graph attention network(GAT)is utilized forfinal feature matching,spanning across the image pair with geometric priors informed by GD.Comprehensive experiments show that GeoGlue outperforms other state-of-the-art methods in feature-matching accuracy and performance stability,achieving pose estimation with maximum rotation and translation errors under 1%in challenging scenes from benchmark datasets,Tanks&Temples and ETH3D.This study also proposes thefirst self-supervised deep-learning model for curved line detection,generating geometric priors for matching so that more attention is put on prominent features and improving the visual effect of 3D reconstruction.展开更多
Traditional Chinese medicine(TCM)is an interesting research topic in China’s thousands of years of history.With the recent advances in artificial intelligence technology,some researchers have started to focus on lear...Traditional Chinese medicine(TCM)is an interesting research topic in China’s thousands of years of history.With the recent advances in artificial intelligence technology,some researchers have started to focus on learning the TCM prescriptions in a data-driven manner.This involves appropriately recommending a set of herbs based on patients’symptoms.Most existing herb recommendation models disregard TCM domain knowledge,for example,the interactions between symptoms and herbs and the TCM-informed observations(i.e.,TCM formulation of prescriptions).In this paper,we propose a knowledge-guided and TCM-informed approach for herb recommendation.The knowledge used includes path interactions and co-occurrence relationships among symptoms and herbs from a knowledge graph generated from TCM literature and prescriptions.The aforementioned knowledge is used to obtain the discriminative feature vectors of symptoms and herbs via a graph attention network.To increase the ability of herb prediction for the given symptoms,we introduce TCM-informed observations in the prediction layer.We apply our proposed model on a TCM prescription dataset,demonstrating significant improvements over state-of-the-art herb recommendation methods.展开更多
Ensuring the safe and efficient operation of self-driving vehicles relies heavily on accurately predicting their future trajectories.Existing approaches commonly employ an encoder-decoder neural network structure to e...Ensuring the safe and efficient operation of self-driving vehicles relies heavily on accurately predicting their future trajectories.Existing approaches commonly employ an encoder-decoder neural network structure to enhance information extraction during the encoding phase.However,these methods often neglect the inclusion of road rule constraints during trajectory formulation in the decoding phase.This paper proposes a novel method that combines neural networks and rule-based constraints in the decoder stage to improve trajectory prediction accuracy while ensuring compliance with vehicle kinematics and road rules.The approach separates vehicle trajectories into lateral and longitudinal routes and utilizes conditional variational autoencoder(CVAE)to capture trajectory uncertainty.The evaluation results demonstrate a reduction of 32.4%and 27.6%in the average displacement error(ADE)for predicting the top five and top ten trajectories,respectively,compared to the baseline method.展开更多
Employing the novel Deep Reinforcement Learning approach,this paper addresses the active power corrective control in modern power systems.Seeking to minimize the joint effect engendered by operation cost and blackout ...Employing the novel Deep Reinforcement Learning approach,this paper addresses the active power corrective control in modern power systems.Seeking to minimize the joint effect engendered by operation cost and blackout penalty,this correction strategy focuses on evaluating the robustness and adaptability aspects of the control agent.In Part I of this paper,where robustness is the primary focus,the agent is developed to handle unexpected incidents and guide the stable operation of power grids A Simulation-driven Graph Attention Reinforcement Learning method is proposed to perform robust active power corrective control.The aim of the graph attention networks is to determine the representation of power system states considering the topological features.Monte Carlo tree search is adopted to select the best suitable action set out of the large action space,including generator redispatch and topology control actions.Finally,driven by simulation,a guided training mechanism along with a long-short-term action deployment strategy are designed to help the agent better evaluate the action set while training and to operate more stably when deployed.The efficacy of the proposed method has been demonstrated in the“2020 I earning to Run a Power Network.Neurips Track 1”global competition and the associated cases.Part II of this paper deals with the adaptability case,where the agent is equipped to better adapt to a grid that has an increasing share of renewable energies through the years.展开更多
基金supported by the Science and Technology Project of State Grid Corporation of China(4000-202122070A-0-0-00).
文摘The fluctuation of wind power affects the operating safety and power consumption of the electric power grid and restricts the grid connection of wind power on a large scale.Therefore,wind power forecasting plays a key role in improving the safety and economic benefits of the power grid.This paper proposes a wind power predicting method based on a convolutional graph attention deep neural network with multi-wind farm data.Based on the graph attention network and attention mechanism,the method extracts spatial-temporal characteristics from the data of multiple wind farms.Then,combined with a deep neural network,a convolutional graph attention deep neural network model is constructed.Finally,the model is trained with the quantile regression loss function to achieve the wind power deterministic and probabilistic prediction based on multi-wind farm spatial-temporal data.A wind power dataset in the U.S.is taken as an example to demonstrate the efficacy of the proposed model.Compared with the selected baseline methods,the proposed model achieves the best prediction performance.The point prediction errors(i.e.,root mean square error(RMSE)and normalized mean absolute percentage error(NMAPE))are 0.304 MW and 1.177%,respectively.And the comprehensive performance of probabilistic prediction(i.e.,con-tinuously ranked probability score(CRPS))is 0.580.Thus,the significance of multi-wind farm data and spatial-temporal feature extraction module is self-evident.
基金supported by the Key Research&Development Plan Project of Shandong Province,China(No.2017GGX10127).
文摘Continuous sign language recognition(CSLR)is challenging due to the complexity of video background,hand gesture variability,and temporal modeling difficulties.This work proposes a CSLR method based on a spatialtemporal graph attention network to focus on essential features of video series.The method considers local details of sign language movements by taking the information on joints and bones as inputs and constructing a spatialtemporal graph to reflect inter-frame relevance and physical connections between nodes.The graph-based multihead attention mechanism is utilized with adjacent matrix calculation for better local-feature exploration,and short-term motion correlation modeling is completed via a temporal convolutional network.We adopted BLSTM to learn the long-termdependence and connectionist temporal classification to align the word-level sequences.The proposed method achieves competitive results regarding word error rates(1.59%)on the Chinese Sign Language dataset and the mean Jaccard Index(65.78%)on the ChaLearn LAP Continuous Gesture Dataset.
文摘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.
基金supported byNationalNatural Science Foundation of China(52274205)and Project of Education Department of Liaoning Province(LJKZ0338).
文摘Automatic text summarization(ATS)plays a significant role in Natural Language Processing(NLP).Abstractive summarization produces summaries by identifying and compressing the most important information in a document.However,there are only relatively several comprehensively evaluated abstractive summarization models that work well for specific types of reports due to their unstructured and oral language text characteristics.In particular,Chinese complaint reports,generated by urban complainers and collected by government employees,describe existing resident problems in daily life.Meanwhile,the reflected problems are required to respond speedily.Therefore,automatic summarization tasks for these reports have been developed.However,similar to traditional summarization models,the generated summaries still exist problems of informativeness and conciseness.To address these issues and generate suitably informative and less redundant summaries,a topic-based abstractive summarization method is proposed to obtain global and local features.Additionally,a heterogeneous graph of the original document is constructed using word-level and topic-level features.Experiments and analyses on public review datasets(Yelp and Amazon)and our constructed dataset(Chinese complaint reports)show that the proposed framework effectively improves the performance of the abstractive summarization model for Chinese complaint reports.
基金Funder One,National Nature Science Foundation of China,Grant/Award No.61972357Funder Two,National Nature Science Foundation of China,Grant/Award No.61672337Funder Three,Guangxi Colleges and Universities Basic Ability Improvement Project of Young and Middle-Aged Teachers,Grant/Award No.2018KY0651.
文摘With the continuous improvement of the e-commerce ecosystem and the rapid growth of e-commerce data, inthe context of the e-commerce ecosystem, consumers ask hundreds of millions of questions every day. In order toimprove the timeliness of customer service responses, many systems have begun to use customer service robotsto respond to consumer questions, but the current customer service robots tend to respond to specific questions.For many questions that lack background knowledge, they can generate only responses that are biased towardsgenerality and repetitiveness. To better promote the understanding of dialogue and generate more meaningfulresponses, this paper introduces knowledge information into the research of question answering system by usinga knowledge graph. The unique structured knowledge base of the knowledge graph is convenient for knowledgequery, can acquire knowledge faster, and improves the background information needed for answering questions. Toavoid the lack of information in the dialogue process, this paper proposes the Multi-hop Knowledge InformationEnhanced Dialogue-Graph Attention (MKIED-GA) model. The model first retrieves the problem subgraph directlyrelated to the input information from the entire knowledge base and then uses the graph neural network as theknowledge inference module on the subgraph to encode the subgraph. The graph attention mechanism is usedto determine the one-hop and two-hop entities that are more relevant to the problem to achieve the aggregationof highly relevant neighbor information. This further enriches the semantic information to provide a betterunderstanding of the meaning of the input question and generate appropriate response information. In the processof generating a response, a multi-attention flow mechanism is used to focus on different information to promotethe generation of better responses. Experiments have proved that the model presented in this article can generatemore meaningful responses than other models.
基金Authors The research project is partially supported by National Natural ScienceFoundation of China under Grant No. 62072015, U19B2039, U1811463National Key R&D Programof China 2018YFB1600903.
文摘PM2.5 concentration prediction is of great significance to environmental protection and human health.Achieving accurate prediction of PM2.5 concentration has become an important research task.However,PM2.5 pollutants can spread in the earth’s atmosphere,causing mutual influence between different cities.To effectively capture the air pollution relationship between cities,this paper proposes a novel spatiotemporal model combining graph attention neural network(GAT)and gated recurrent unit(GRU),named GAT-GRU for PM2.5 concentration prediction.Specifically,GAT is used to learn the spatial dependence of PM2.5 concentration data in different cities,and GRU is to extract the temporal dependence of the long-term data series.The proposed model integrates the learned spatio-temporal dependencies to capture long-term complex spatio-temporal features.Considering that air pollution is related to the meteorological conditions of the city,the knowledge acquired from meteorological data is used in the model to enhance PM2.5 prediction performance.The input of the GAT-GRU model consists of PM2.5 concentration data and meteorological data.In order to verify the effectiveness of the proposed GAT-GRU prediction model,this paper designs experiments on real-world datasets compared with other baselines.Experimental results prove that our model achieves excellent performance in PM2.5 concentration prediction.
基金This work was supported by the National Key Research and Development Program of China under Grant No.2021YFB3100600the Youth Innovation Promotion Association of Chinese Academy of Sciences under Grant No.2021153the State Key Program of National Natural Science Foundation of China under Grant No.U2336202.
文摘Event detection(ED)seeks to recognize event triggers and classify them into the predefined event types.Chinese ED is formulated as a character-level task owing to the uncertain word boundaries.Prior methods try to incorpo-rate word-level information into characters to enhance their semantics.However,they experience two problems.First,they fail to incorporate word-level information into each character the word encompasses,causing the insufficient word-charac-ter interaction problem.Second,they struggle to distinguish events of similar types with limited annotated instances,which is called the event confusing problem.This paper proposes a novel model named Label-Aware Heterogeneous Graph Attention Network(L-HGAT)to address these two problems.Specifically,we first build a heterogeneous graph of two node types and three edge types to maximally preserve word-character interactions,and then deploy a heterogeneous graph attention network to enhance the semantic propagation between characters and words.Furthermore,we design a pushing-away game to enlarge the predicting gap between the ground-truth event type and its confusing counterpart for each character.Experimental results show that our L-HGAT model consistently achieves superior performance over prior competitive methods.
基金the National Natural Science Foundation of China(Grant Nos.11804076 and 91961204)the Fundamental Research Funds for the Central Universities of China(No.B210202151)the Changzhou Science and Technology Plan(No.CZ520012712).
文摘Due to coexistence of huge number of structural isomers,global search for the ground-state structures of atomic clusters is a challenging issue.The difficulty also originates from the computational cost of ab initio methods for describing the potential energy surface.Recently,machine learning techniques have been widely utilized to accelerate materials discovery and molecular simulation.Compared to the commonly used artificial neural network,graph network is naturally suitable for clusters with flexible geometric environment of each atom.Herein we develop a cluster graph attention network(CGANet)by aggregating information of neighboring vertices and edges using attention mechanism,which can precisely predict the binding energy and force of silver clusters with root mean square error of 5.4 meV/atom and mean absolute error of 42.3 meV/Å,respectively.As a proof-of-concept,we have performed global optimization of mediumsized Agn clusters(n=14–26)by combining CGANet and genetic algorithm.The reported ground-state structures for n=14–21,have been successfully reproduced,while entirely new lowest-energy structures are obtained for n=22–26.In addition to the description of potential energy surface,the CGANet is also applied to predict the electronic properties of clusters,such as HOMO energy and HOMO-LUMO gap.With accuracy comparable to ab initio methods and acceleration by at least two orders of magnitude,CGANet holds great promise in global search of lowest-energy structures of large clusters and inverse design of functional clusters.
基金supported by National Natural Science Foundation of China(62003028).X.L.was supported by a Scholarship from the China Scholarship Council.
文摘Recent advances in spatially resolved transcriptomic technologies have enabled unprecedented opportunities to elucidate tissue architecture and function in situ.Spatial transcriptomics can provide multimodal and complementary information simultaneously,including gene expression profiles,spatial locations,and histology images.However,most existing methods have limitations in efficiently utilizing spatial information and matched high-resolution histology images.To fully leverage the multi-modal information,we propose a SPAtially embedded Deep Attentional graph Clustering(SpaDAC)method to identify spatial domains while reconstructing denoised gene expression profiles.This method can efficiently learn the low-dimensional embeddings for spatial transcriptomics data by constructing multi-view graph modules to capture both spatial location connectives and morphological connectives.Benchmark results demonstrate that SpaDAC outperforms other algorithms on several recent spatial transcriptomics datasets.SpaDAC is a valuable tool for spatial domain detection,facilitating the comprehension of tissue architecture and cellular microenvironment.The source code of SpaDAC is freely available at Github(https://github.com/huoyuying/SpaDAC.git).
基金This work was supported in part by the National Natural Science Foundation of China under Grants 62273272,62303375 and 61873277in part by the Key Research and Development Program of Shaanxi Province under Grant 2023-YBGY-243+2 种基金in part by the Natural Science Foundation of Shaanxi Province under Grants 2022JQ-606 and 2020-JQ758in part by the Research Plan of Department of Education of Shaanxi Province under Grant 21JK0752in part by the Youth Innovation Team of Shaanxi Universities.
文摘Social robot accounts controlled by artificial intelligence or humans are active in social networks,bringing negative impacts to network security and social life.Existing social robot detection methods based on graph neural networks suffer from the problem of many social network nodes and complex relationships,which makes it difficult to accurately describe the difference between the topological relations of nodes,resulting in low detection accuracy of social robots.This paper proposes a social robot detection method with the use of an improved neural network.First,social relationship subgraphs are constructed by leveraging the user’s social network to disentangle intricate social relationships effectively.Then,a linear modulated graph attention residual network model is devised to extract the node and network topology features of the social relation subgraph,thereby generating comprehensive social relation subgraph features,and the feature-wise linear modulation module of the model can better learn the differences between the nodes.Next,user text content and behavioral gene sequences are extracted to construct social behavioral features combined with the social relationship subgraph features.Finally,social robots can be more accurately identified by combining user behavioral and relationship features.By carrying out experimental studies based on the publicly available datasets TwiBot-20 and Cresci-15,the suggested method’s detection accuracies can achieve 86.73%and 97.86%,respectively.Compared with the existing mainstream approaches,the accuracy of the proposed method is 2.2%and 1.35%higher on the two datasets.The results show that the method proposed in this paper can effectively detect social robots and maintain a healthy ecological environment of social networks.
基金the National Natural Science Foundation of China(Grant Nos.11861045,11361033,and 62162040)。
文摘Essential proteins play an important role in disease diagnosis and drug development.Many methods have been devoted to the essential protein prediction by using some kinds of biological information.However,they either ignore the noise presented in the biological information itself or the noise generated during feature extraction.To overcome these problems,in this paper,we propose a novel method for predicting essential proteins called attention gate-graph attention network and temporal convolutional network(AG-GATCN).In AG-GATCN method,we use improved temporal convolutional network(TCN)to extract features from gene expression sequence.To address the noise in the gene expression sequence itself and the noise generated after the dilated causal convolution,we introduce attention mechanism and gating mechanism in TCN.In addition,we use graph attention network(GAT)to extract protein–protein interaction(PPI)network features,in which we construct the feature matrix by introducing node2vec technique and 7 centrality metrics,and to solve the GAT oversmoothing problem,we introduce gated tanh unit(GTU)in GAT.Finally,two types of features are integrated by us to predict essential proteins.Compared with the existing methods for predicting essential proteins,the experimental results show that AG-GATCN achieves better performance.
基金supported in part by the National Natural Science Foundation of China under Grant 62272062the Researchers Supporting Project number.(RSP2023R102)King Saud University+5 种基金Riyadh,Saudi Arabia,the Open Research Fund of the Hunan Provincial Key Laboratory of Network Investigational Technology under Grant 2018WLZC003the National Science Foundation of Hunan Province under Grant 2020JJ2029the Hunan Provincial Key Research and Development Program under Grant 2022GK2019the Science Fund for Creative Research Groups of Hunan Province under Grant 2020JJ1006the Scientific Research Fund of Hunan Provincial Transportation Department under Grant 202143the Open Fund of Key Laboratory of Safety Control of Bridge Engineering,Ministry of Education(Changsha University of Science Technology)under Grant 21KB07.
文摘Sensors produce a large amount of multivariate time series data to record the states of Internet of Things(IoT)systems.Multivariate time series timestamp anomaly detection(TSAD)can identify timestamps of attacks and malfunctions.However,it is necessary to determine which sensor or indicator is abnormal to facilitate a more detailed diagnosis,a process referred to as fine-grained anomaly detection(FGAD).Although further FGAD can be extended based on TSAD methods,existing works do not provide a quantitative evaluation,and the performance is unknown.Therefore,to tackle the FGAD problem,this paper first verifies that the TSAD methods achieve low performance when applied to the FGAD task directly because of the excessive fusion of features and the ignoring of the relationship’s dynamic changes between indicators.Accordingly,this paper proposes a mul-tivariate time series fine-grained anomaly detection(MFGAD)framework.To avoid excessive fusion of features,MFGAD constructs two sub-models to independently identify the abnormal timestamp and abnormal indicator instead of a single model and then combines the two kinds of abnormal results to detect the fine-grained anomaly.Based on this framework,an algorithm based on Graph Attention Neural Network(GAT)and Attention Convolutional Long-Short Term Memory(A-ConvLSTM)is proposed,in which GAT learns temporal features of multiple indicators to detect abnormal timestamps and A-ConvLSTM captures the dynamic relationship between indicators to identify abnormal indicators.Extensive simulations on a real-world dataset demonstrate that the proposed algorithm can achieve a higher F1 score and hit rate than the extension of existing TSAD methods with the benefit of two independent sub-models for timestamp and indicator detection.
基金the National Natural Science Foundation of China(NNSFC)(Grant Nos.72001213 and 72301292)the National Social Science Fund of China(Grant No.19BGL297)the Basic Research Program of Natural Science in Shaanxi Province(Grant No.2021JQ-369).
文摘Due to the time-varying topology and possible disturbances in a conflict environment,it is still challenging to maintain the mission performance of flying Ad hoc networks(FANET),which limits the application of Unmanned Aerial Vehicle(UAV)swarms in harsh environments.This paper proposes an intelligent framework to quickly recover the cooperative coveragemission by aggregating the historical spatio-temporal network with the attention mechanism.The mission resilience metric is introduced in conjunction with connectivity and coverage status information to simplify the optimization model.A spatio-temporal node pooling method is proposed to ensure all node location features can be updated after destruction by capturing the temporal network structure.Combined with the corresponding Laplacian matrix as the hyperparameter,a recovery algorithm based on the multi-head attention graph network is designed to achieve rapid recovery.Simulation results showed that the proposed framework can facilitate rapid recovery of the connectivity and coverage more effectively compared to the existing studies.The results demonstrate that the average connectivity and coverage results is improved by 17.92%and 16.96%,respectively compared with the state-of-the-art model.Furthermore,by the ablation study,the contributions of each different improvement are compared.The proposed model can be used to support resilient network design for real-time mission execution.
基金supported by the National key Research&Development Program of China(No.2017YFC0820503)the National Natural Science Foundation of China(No.62072149)+2 种基金the National Social Science Foundation of China(No.19ZDA348)the Primary Research&Development Plan of Zhejiang(No.2021C03156)the Public Welfare Research Program of Zhejiang(No.LGG19F020017)。
文摘Event temporal relation extraction is an important part of natural language processing.Many models are being used in this task with the development of deep learning.However,most of the existing methods cannot accurately obtain the degree of association between different tokens and events,and event-related information cannot be effectively integrated.In this paper,we propose an event information integration model that integrates event information through multilayer bidirectional long short-term memory(Bi-LSTM)and attention mechanism.Although the above scheme can improve the extraction performance,it can still be further optimized.To further improve the performance of the previous scheme,we propose a novel relational graph attention network that incorporates edge attributes.In this approach,we first build a semantic dependency graph through dependency parsing,model a semantic graph that considers the edges’attributes by using top-k attention mechanisms to learn hidden semantic contextual representations,and finally predict event temporal relations.We evaluate proposed models on the TimeBank-Dense dataset.Compared to previous baselines,the Micro-F1 scores obtained by our models improve by 3.9%and 14.5%,respectively.
基金This work is supported by the National Natural Science Foundation of China(61976028,61572085,61806026,61502058).
文摘For a long time, the detection and extraction of printed surfacedefects has been a hot issue in the print industry. Nowadays, defect detectionof a large number of products still relies on traditional image processingalgorithms such as scale invariant feature transform (SIFT) and orientedfast and rotated brief (ORB), and researchers need to design algorithms forspecific products. At present, a large number of defect detection algorithmsbased on object detection have been applied but need lots of labeling sampleswith defects. Besides, there are many kinds of defects in printed surface,so it is difficult to enumerate all defects. Most defect detection based onunsupervised learning of positive samples use generative adversarial networks(GAN) and variational auto-encoders (VAE) algorithms, but these methodsare not effective for complex printed surface. Aiming at these problems, Inthis paper, an unsupervised defect detection and extraction algorithm forprinted surface based on positive samples in the complex printed surface isproposed innovatively. We propose a kind of defect detection and extractionnetwork based on image matching network. This network is divided into thefull convolution network of feature points extraction, and the graph attentionnetwork using self attention and cross attention. Though the key pointsextraction network, we can get robustness key points in the complex printedimages, and the graph network can solve the problem of the deviation becauseof different camera positions and the influence of defect in the differentproduction lines. Just one positive sample image is needed as the benchmarkto detect the defects. The algorithm in this paper has been proved in “TheFirst ZhengTu Cup on Campus Machine Vision AI Competition” and gotexcellent results in the finals. We are working with the company to apply it inproduction.
基金supported by the Strategic Priority Research Program of the Chinese Academy of Sciences[Grant No.XDA19080101]the Director Fund of the International Research Center of Big Data for Sus-tainable Development Goals[Grant No.CBAS2022DF015]+1 种基金the National Natural Science Foundation of China[Grant No.41901328 and 41974108]the National Key Research and Development Program of China[Grant No.2022YFC3800700].
文摘We present GeoGlue,a novel method using high-resolution UAV imagery for accurate feature matching,which is normally challenging due to the complicated scenes.Current feature detection methods are performed without guidance of geometric priors(e.g.,geometric lines),lacking enough attention given to salient geometric features which are indispensable for accurate matching due to their stable existence across views.In this work,geometric lines arefirstly detected by a CNN-based geometry detector(GD)which is pre-trained in a self-supervised manner through automatically generated images.Then,geometric lines are naturally vectorized based on GD and thus non-significant features can be disregarded as judged by their disordered geometric morphology.A graph attention network(GAT)is utilized forfinal feature matching,spanning across the image pair with geometric priors informed by GD.Comprehensive experiments show that GeoGlue outperforms other state-of-the-art methods in feature-matching accuracy and performance stability,achieving pose estimation with maximum rotation and translation errors under 1%in challenging scenes from benchmark datasets,Tanks&Temples and ETH3D.This study also proposes thefirst self-supervised deep-learning model for curved line detection,generating geometric priors for matching so that more attention is put on prominent features and improving the visual effect of 3D reconstruction.
基金supported by the China Knowledge Centre for Engi-neering Sciences and Technology(CKCEST)and the National Natural Science Foundation of China(No.62037001)。
文摘Traditional Chinese medicine(TCM)is an interesting research topic in China’s thousands of years of history.With the recent advances in artificial intelligence technology,some researchers have started to focus on learning the TCM prescriptions in a data-driven manner.This involves appropriately recommending a set of herbs based on patients’symptoms.Most existing herb recommendation models disregard TCM domain knowledge,for example,the interactions between symptoms and herbs and the TCM-informed observations(i.e.,TCM formulation of prescriptions).In this paper,we propose a knowledge-guided and TCM-informed approach for herb recommendation.The knowledge used includes path interactions and co-occurrence relationships among symptoms and herbs from a knowledge graph generated from TCM literature and prescriptions.The aforementioned knowledge is used to obtain the discriminative feature vectors of symptoms and herbs via a graph attention network.To increase the ability of herb prediction for the given symptoms,we introduce TCM-informed observations in the prediction layer.We apply our proposed model on a TCM prescription dataset,demonstrating significant improvements over state-of-the-art herb recommendation methods.
基金supported in part by the National Natural Science Foundation of China under Grant 52372393,62003238in part by the DongfengTechnology Center(Research and Application of Next-Generation Low-Carbonntelligent Architecture Technology).
文摘Ensuring the safe and efficient operation of self-driving vehicles relies heavily on accurately predicting their future trajectories.Existing approaches commonly employ an encoder-decoder neural network structure to enhance information extraction during the encoding phase.However,these methods often neglect the inclusion of road rule constraints during trajectory formulation in the decoding phase.This paper proposes a novel method that combines neural networks and rule-based constraints in the decoder stage to improve trajectory prediction accuracy while ensuring compliance with vehicle kinematics and road rules.The approach separates vehicle trajectories into lateral and longitudinal routes and utilizes conditional variational autoencoder(CVAE)to capture trajectory uncertainty.The evaluation results demonstrate a reduction of 32.4%and 27.6%in the average displacement error(ADE)for predicting the top five and top ten trajectories,respectively,compared to the baseline method.
基金supported by the National Key R&D Program of China under Grant 2018AAA0101504the Science and technology project of SGCC(State Grid Corporation of China):fundamental theory of human-in-the-oop hybrid-augmented intelligence for power grid dispatch and control.
文摘Employing the novel Deep Reinforcement Learning approach,this paper addresses the active power corrective control in modern power systems.Seeking to minimize the joint effect engendered by operation cost and blackout penalty,this correction strategy focuses on evaluating the robustness and adaptability aspects of the control agent.In Part I of this paper,where robustness is the primary focus,the agent is developed to handle unexpected incidents and guide the stable operation of power grids A Simulation-driven Graph Attention Reinforcement Learning method is proposed to perform robust active power corrective control.The aim of the graph attention networks is to determine the representation of power system states considering the topological features.Monte Carlo tree search is adopted to select the best suitable action set out of the large action space,including generator redispatch and topology control actions.Finally,driven by simulation,a guided training mechanism along with a long-short-term action deployment strategy are designed to help the agent better evaluate the action set while training and to operate more stably when deployed.The efficacy of the proposed method has been demonstrated in the“2020 I earning to Run a Power Network.Neurips Track 1”global competition and the associated cases.Part II of this paper deals with the adaptability case,where the agent is equipped to better adapt to a grid that has an increasing share of renewable energies through the years.