With the popularity of 5G and the rapid development of mobile terminals,an endless stream of short video software exists.Browsing short-form mobile video in fragmented time has become the mainstream of user’s life.He...With the popularity of 5G and the rapid development of mobile terminals,an endless stream of short video software exists.Browsing short-form mobile video in fragmented time has become the mainstream of user’s life.Hence,designing an efficient short video recommendation method has become important for major network platforms to attract users and satisfy their requirements.Nevertheless,the explosive growth of data leads to the low efficiency of the algorithm,which fails to distill users’points of interest on one hand effectively.On the other hand,integrating user preferences and the content of items urgently intensify the requirements for platform recommendation.In this paper,we propose a collaborative filtering algorithm,integrating time context information and user context,which pours attention into expanding and discovering user interest.In the first place,we introduce the temporal context information into the typical collaborative filtering algorithm,and leverage the popularity penalty function to weight the similarity between recommended short videos and the historical short videos.There remains one more point.We also introduce the user situation into the traditional collaborative filtering recommendation algorithm,considering the context information of users in the generation recommendation stage,and weight the recommended short-formvideos of candidates.At last,a diverse approach is used to generate a Top-K recommendation list for users.And through a case study,we illustrate the accuracy and diversity of the proposed method.展开更多
With the improvement of the national economic level,the number of vehicles is still increasing year by year.According to the statistics of National Bureau of Statics,the number is approximately up to 327 million in Ch...With the improvement of the national economic level,the number of vehicles is still increasing year by year.According to the statistics of National Bureau of Statics,the number is approximately up to 327 million in China by the end of 2018,which makes urban traffic pressure continues to rise so that the negative impact of urban traffic order is growing.Illegal parking-the common problem in the field of transportation security is urgent to be solved and traditional methods to address it are mainly based on ground loop and manual supervision,which may miss detection and cost much manpower.Due to the rapidly developing deep learning sweeping the world in recent years,object detection methods relying on background segmentation cannot meet the requirements of complex and various scenes on speed and precision.Thus,an improved Single Shot MultiBox Detector(SSD)based on deep learning is proposed in our study,we introduce attention mechanism by spatial transformer module which gives neural networks the ability to actively spatially transform feature maps and add contextual information transmission in specified layer.Finally,we found out the best connection layer in the detection model by repeated experiments especially for small objects and increased the precision by 1.5%than the baseline SSD without extra training cost.Meanwhile,we designed an illegal parking vehicle detection method by the improved SSD,reaching a high precision up to 97.3%and achieving a speed of 40FPS,superior to most of vehicle detection methods,will make contributions to relieving the negative impact of illegal parking.展开更多
Understanding an image goes beyond recognizing and locating the objects in it,the relationships between objects also very important in image understanding.Most previous methods have focused on recognizing local predic...Understanding an image goes beyond recognizing and locating the objects in it,the relationships between objects also very important in image understanding.Most previous methods have focused on recognizing local predictions of the relationships.But real-world image relationships often determined by the surrounding objects and other contextual information.In this work,we employ this insight to propose a novel framework to deal with the problem of visual relationship detection.The core of the framework is a relationship inference network,which is a recurrent structure designed for combining the global contextual information of the object to infer the relationship of the image.Experimental results on Stanford VRD and Visual Genome demonstrate that the proposed method achieves a good performance both in efficiency and accuracy.Finally,we demonstrate the value of visual relationship on two computer vision tasks:image retrieval and scene graph generation.展开更多
Aspect-based sentiment analysis aims to detect and classify the sentiment polarities as negative,positive,or neutral while associating them with their identified aspects from the corresponding context.In this regard,p...Aspect-based sentiment analysis aims to detect and classify the sentiment polarities as negative,positive,or neutral while associating them with their identified aspects from the corresponding context.In this regard,prior methodologies widely utilize either word embedding or tree-based rep-resentations.Meanwhile,the separate use of those deep features such as word embedding and tree-based dependencies has become a significant cause of information loss.Generally,word embedding preserves the syntactic and semantic relations between a couple of terms lying in a sentence.Besides,the tree-based structure conserves the grammatical and logical dependencies of context.In addition,the sentence-oriented word position describes a critical factor that influences the contextual information of a targeted sentence.Therefore,knowledge of the position-oriented information of words in a sentence has been considered significant.In this study,we propose to use word embedding,tree-based representation,and contextual position information in combination to evaluate whether their combination will improve the result’s effectiveness or not.In the meantime,their joint utilization enhances the accurate identification and extraction of targeted aspect terms,which also influences their classification process.In this research paper,we propose a method named Attention Based Multi-Channel Convolutional Neural Net-work(Att-MC-CNN)that jointly utilizes these three deep features such as word embedding with tree-based structure and contextual position informa-tion.These three parameters deliver to Multi-Channel Convolutional Neural Network(MC-CNN)that identifies and extracts the potential terms and classifies their polarities.In addition,these terms have been further filtered with the attention mechanism,which determines the most significant words.The empirical analysis proves the proposed approach’s effectiveness compared to existing techniques when evaluated on standard datasets.The experimental results represent our approach outperforms in the F1 measure with an overall achievement of 94%in identifying aspects and 92%in the task of sentiment classification.展开更多
Medical image segmentation plays a crucial role in clinical diagnosis and therapy systems,yet still faces many challenges.Building on convolutional neural networks(CNNs),medical image segmentation has achieved tremend...Medical image segmentation plays a crucial role in clinical diagnosis and therapy systems,yet still faces many challenges.Building on convolutional neural networks(CNNs),medical image segmentation has achieved tremendous progress.However,owing to the locality of convolution operations,CNNs have the inherent limitation in learning global context.To address the limitation in building global context relationship from CNNs,we propose LGNet,a semantic segmentation network aiming to learn local and global features for fast and accurate medical image segmentation in this paper.Specifically,we employ a two-branch architecture consisting of convolution layers in one branch to learn local features and transformer layers in the other branch to learn global features.LGNet has two key insights:(1)We bridge two-branch to learn local and global features in an interactive way;(2)we present a novel multi-feature fusion model(MSFFM)to leverage the global contexture information from transformer and the local representational features from convolutions.Our method achieves state-of-the-art trade-off in terms of accuracy and efficiency on several medical image segmentation benchmarks including Synapse,ACDC and MOST.Specifically,LGNet achieves the state-of-the-art performance with Dice's indexes of 80.15%on Synapse,of 91.70%on ACDC,and of 95.56%on MOST.Meanwhile,the inference speed attains at 172 frames per second with 224-224 input resolution.The extensive experiments demonstrate the effectiveness of the proposed LGNet for fast and accurate for medical image segmentation.展开更多
Automatic bridge detection is an important application of SAR images. Differed from the classical CFAR method, a new knowledge-based bridge detection approach is proposed. The method not only uses the backscattering i...Automatic bridge detection is an important application of SAR images. Differed from the classical CFAR method, a new knowledge-based bridge detection approach is proposed. The method not only uses the backscattering intensity difference between targets and background but also applies the contextual information and spatial relationship between objects. According to bridges' special characteristics and scattering properties in SAR images, the new knowledge-based method includes three processes: river segmentation, potential bridge areas detection and bridge discrimination. The application to AIRSAR data shows that the new method is not sensitive to rivers' shape. Moreover, this method can detect bridges successfully when river segmentation is not very exact and is more robust than the radius projection method.展开更多
As cloud service becomes more and more capable, available and powerful, wiseCIO has emerged from an innovative roadmap toward archival Content Management Service (aCMS) and massive Content Delivery Service (mCDS) in s...As cloud service becomes more and more capable, available and powerful, wiseCIO has emerged from an innovative roadmap toward archival Content Management Service (aCMS) and massive Content Delivery Service (mCDS) in support of Anything-as-a-Service (XaaS) via Digital Archiving and Transformed Analytics (DATA);DATA aims to automate UBC with FAST solutions throughout a feasible, analytical, scalable and testable approach. This paper, based on the novel wiseCIO (web-based intelligent service engaging Cloud Intelligence Outlet), presents digital archiving and transformed analytics via machine learning automata for intelligent UBC processes to liaise with Universal interface for human-computer interaction, enable Brewing aggregation (differing from traditional web browsing), and engage Centered user experience. As one of the most practical aspects of artificial intelligence, machine learning is applied to analytical model building and massive and/or multidimensional Online Analytical Processing (mOLAP) for more intelligent cloud service with little explicit coding required. DATA is central to useful information via archival transformation and analytics, and utilizable intelligence for Business, Education and Entertainment (iBEE) in support of decision-making. As a result, DATA orchestrates wiseCIO to promote ACTiVE XaaS that enables accessibility, contextuality and traceability of information for vast engagement with various cloud services, such as aCMS (archival Content Management Service), COSA (Context-Oriented Screening Aggregation), DASH (Deliveries Assembled for fast Search and Hits), OLAS (Online Learning via Analytical Synthesis), REAP (Rapid Extension and Active Presentation), and SPOT (Special Points On Top) with great ease.展开更多
Self-attention aggregates similar feature information to enhance the features. However, the attention covers nonface areas in face alignment, which may be disturbed in challenging cases, such as occlusions, and fails ...Self-attention aggregates similar feature information to enhance the features. However, the attention covers nonface areas in face alignment, which may be disturbed in challenging cases, such as occlusions, and fails to predict landmarks. In addition, the learned feature similarity variance is not large enough in the experiment. To this end, we propose structural dependence learning based on self-attention for face alignment (SSFA). It limits the self-attention learning to the facial range and adaptively builds the significant landmark structure dependency. Compared with other state-of-the-art methods, SSFA effectively improves the performance on several standard facial landmark detection benchmarks and adapts more in challenging cases.展开更多
Database system is the infrastructure of the modern information system. The R&D in the database system and its technologies is one of the important research topics in the field. The database R&D in China took off la...Database system is the infrastructure of the modern information system. The R&D in the database system and its technologies is one of the important research topics in the field. The database R&D in China took off later but it moves along by giant steps. This report presents the achievements Renmin University of China (RUC) has made in the past 25 years and at the same time addresses some of the research projects we, RUC, are currently working on. The National Natural Science Foundation of China supports and initiates most of our research projects and these successfully conducted projects have produced fruitful results.展开更多
Recommendation systems have been extensively studied over the last decade in various domains. It has been considered a powerful tool for assisting business owners in promoting sales and helping users with decision-mak...Recommendation systems have been extensively studied over the last decade in various domains. It has been considered a powerful tool for assisting business owners in promoting sales and helping users with decision-making when given numerous choices. In this paper, we propose a novel Graph-based Context-Aware Recommendation Systems with Knowledge Graph to analyse and predict users’ behaviours, i.e., making recommendations based on historical events and their implicit associations. The model incorporates contextual information extracted from both users’ historical behaviours and events relations, where the contexts have been modelled as knowledge graphs. By leveraging the advantages offered from the knowledge graph, events dependencies and their subtle relations can be established and have been introduced in the recommendation process. Experimental results indicate that the proposed approach can outperform the state-of-the-art algorithms and achieve more accurate recommendations.展开更多
Survey generation aims to generate a summary from a scientific topic based on related papers.The structure of papers deeply influences the generative process of survey,especially the relationships between sentence and...Survey generation aims to generate a summary from a scientific topic based on related papers.The structure of papers deeply influences the generative process of survey,especially the relationships between sentence and sentence,paragraph and paragraph.In principle,the structure of paper can influence the quality of the summary.Therefore,we employ the structure of paper to leverage contextual information among sentences in paragraphs to generate a survey for documents.In particular,we present a neural document structure model for survey generation.We take paragraphs as units,and model sentences in paragraphs,we then employ a hierarchical model to learn structure among sentences,which can be used to select important and informative sentences to generate survey.We evaluate our model on scientific document data set.The experimental results show that our model is effective,and the generated survey is informative and readable.展开更多
文摘With the popularity of 5G and the rapid development of mobile terminals,an endless stream of short video software exists.Browsing short-form mobile video in fragmented time has become the mainstream of user’s life.Hence,designing an efficient short video recommendation method has become important for major network platforms to attract users and satisfy their requirements.Nevertheless,the explosive growth of data leads to the low efficiency of the algorithm,which fails to distill users’points of interest on one hand effectively.On the other hand,integrating user preferences and the content of items urgently intensify the requirements for platform recommendation.In this paper,we propose a collaborative filtering algorithm,integrating time context information and user context,which pours attention into expanding and discovering user interest.In the first place,we introduce the temporal context information into the typical collaborative filtering algorithm,and leverage the popularity penalty function to weight the similarity between recommended short videos and the historical short videos.There remains one more point.We also introduce the user situation into the traditional collaborative filtering recommendation algorithm,considering the context information of users in the generation recommendation stage,and weight the recommended short-formvideos of candidates.At last,a diverse approach is used to generate a Top-K recommendation list for users.And through a case study,we illustrate the accuracy and diversity of the proposed method.
基金This research has been supported by NSFC(61672495)Scientific Research Fund of Hunan Provincial Education Department(16A208)+1 种基金Project of Hunan Provincial Science and Technology Department(2017SK2405)in part by the construct program of the key discipline in Hunan Province and the CERNET Innovation Project(NGII20170715).
文摘With the improvement of the national economic level,the number of vehicles is still increasing year by year.According to the statistics of National Bureau of Statics,the number is approximately up to 327 million in China by the end of 2018,which makes urban traffic pressure continues to rise so that the negative impact of urban traffic order is growing.Illegal parking-the common problem in the field of transportation security is urgent to be solved and traditional methods to address it are mainly based on ground loop and manual supervision,which may miss detection and cost much manpower.Due to the rapidly developing deep learning sweeping the world in recent years,object detection methods relying on background segmentation cannot meet the requirements of complex and various scenes on speed and precision.Thus,an improved Single Shot MultiBox Detector(SSD)based on deep learning is proposed in our study,we introduce attention mechanism by spatial transformer module which gives neural networks the ability to actively spatially transform feature maps and add contextual information transmission in specified layer.Finally,we found out the best connection layer in the detection model by repeated experiments especially for small objects and increased the precision by 1.5%than the baseline SSD without extra training cost.Meanwhile,we designed an illegal parking vehicle detection method by the improved SSD,reaching a high precision up to 97.3%and achieving a speed of 40FPS,superior to most of vehicle detection methods,will make contributions to relieving the negative impact of illegal parking.
文摘Understanding an image goes beyond recognizing and locating the objects in it,the relationships between objects also very important in image understanding.Most previous methods have focused on recognizing local predictions of the relationships.But real-world image relationships often determined by the surrounding objects and other contextual information.In this work,we employ this insight to propose a novel framework to deal with the problem of visual relationship detection.The core of the framework is a relationship inference network,which is a recurrent structure designed for combining the global contextual information of the object to infer the relationship of the image.Experimental results on Stanford VRD and Visual Genome demonstrate that the proposed method achieves a good performance both in efficiency and accuracy.Finally,we demonstrate the value of visual relationship on two computer vision tasks:image retrieval and scene graph generation.
基金supported by the Deanship of Scientific Research,Vice Presidency for Graduate Studies and Scientific Research,King Faisal University,Saudi Arabia[Grant No.3418].
文摘Aspect-based sentiment analysis aims to detect and classify the sentiment polarities as negative,positive,or neutral while associating them with their identified aspects from the corresponding context.In this regard,prior methodologies widely utilize either word embedding or tree-based rep-resentations.Meanwhile,the separate use of those deep features such as word embedding and tree-based dependencies has become a significant cause of information loss.Generally,word embedding preserves the syntactic and semantic relations between a couple of terms lying in a sentence.Besides,the tree-based structure conserves the grammatical and logical dependencies of context.In addition,the sentence-oriented word position describes a critical factor that influences the contextual information of a targeted sentence.Therefore,knowledge of the position-oriented information of words in a sentence has been considered significant.In this study,we propose to use word embedding,tree-based representation,and contextual position information in combination to evaluate whether their combination will improve the result’s effectiveness or not.In the meantime,their joint utilization enhances the accurate identification and extraction of targeted aspect terms,which also influences their classification process.In this research paper,we propose a method named Attention Based Multi-Channel Convolutional Neural Net-work(Att-MC-CNN)that jointly utilizes these three deep features such as word embedding with tree-based structure and contextual position informa-tion.These three parameters deliver to Multi-Channel Convolutional Neural Network(MC-CNN)that identifies and extracts the potential terms and classifies their polarities.In addition,these terms have been further filtered with the attention mechanism,which determines the most significant words.The empirical analysis proves the proposed approach’s effectiveness compared to existing techniques when evaluated on standard datasets.The experimental results represent our approach outperforms in the F1 measure with an overall achievement of 94%in identifying aspects and 92%in the task of sentiment classification.
基金supported by the Open-Fund of WNLO (Grant No.2018WNLOKF027)the Hubei Key Laboratory of Intelligent Robot in Wuhan Institute of Technology (Grant No.HBIRL 202003).
文摘Medical image segmentation plays a crucial role in clinical diagnosis and therapy systems,yet still faces many challenges.Building on convolutional neural networks(CNNs),medical image segmentation has achieved tremendous progress.However,owing to the locality of convolution operations,CNNs have the inherent limitation in learning global context.To address the limitation in building global context relationship from CNNs,we propose LGNet,a semantic segmentation network aiming to learn local and global features for fast and accurate medical image segmentation in this paper.Specifically,we employ a two-branch architecture consisting of convolution layers in one branch to learn local features and transformer layers in the other branch to learn global features.LGNet has two key insights:(1)We bridge two-branch to learn local and global features in an interactive way;(2)we present a novel multi-feature fusion model(MSFFM)to leverage the global contexture information from transformer and the local representational features from convolutions.Our method achieves state-of-the-art trade-off in terms of accuracy and efficiency on several medical image segmentation benchmarks including Synapse,ACDC and MOST.Specifically,LGNet achieves the state-of-the-art performance with Dice's indexes of 80.15%on Synapse,of 91.70%on ACDC,and of 95.56%on MOST.Meanwhile,the inference speed attains at 172 frames per second with 224-224 input resolution.The extensive experiments demonstrate the effectiveness of the proposed LGNet for fast and accurate for medical image segmentation.
基金supported by the National Key Laboratory of ATR(9140C8002010706).
文摘Automatic bridge detection is an important application of SAR images. Differed from the classical CFAR method, a new knowledge-based bridge detection approach is proposed. The method not only uses the backscattering intensity difference between targets and background but also applies the contextual information and spatial relationship between objects. According to bridges' special characteristics and scattering properties in SAR images, the new knowledge-based method includes three processes: river segmentation, potential bridge areas detection and bridge discrimination. The application to AIRSAR data shows that the new method is not sensitive to rivers' shape. Moreover, this method can detect bridges successfully when river segmentation is not very exact and is more robust than the radius projection method.
文摘As cloud service becomes more and more capable, available and powerful, wiseCIO has emerged from an innovative roadmap toward archival Content Management Service (aCMS) and massive Content Delivery Service (mCDS) in support of Anything-as-a-Service (XaaS) via Digital Archiving and Transformed Analytics (DATA);DATA aims to automate UBC with FAST solutions throughout a feasible, analytical, scalable and testable approach. This paper, based on the novel wiseCIO (web-based intelligent service engaging Cloud Intelligence Outlet), presents digital archiving and transformed analytics via machine learning automata for intelligent UBC processes to liaise with Universal interface for human-computer interaction, enable Brewing aggregation (differing from traditional web browsing), and engage Centered user experience. As one of the most practical aspects of artificial intelligence, machine learning is applied to analytical model building and massive and/or multidimensional Online Analytical Processing (mOLAP) for more intelligent cloud service with little explicit coding required. DATA is central to useful information via archival transformation and analytics, and utilizable intelligence for Business, Education and Entertainment (iBEE) in support of decision-making. As a result, DATA orchestrates wiseCIO to promote ACTiVE XaaS that enables accessibility, contextuality and traceability of information for vast engagement with various cloud services, such as aCMS (archival Content Management Service), COSA (Context-Oriented Screening Aggregation), DASH (Deliveries Assembled for fast Search and Hits), OLAS (Online Learning via Analytical Synthesis), REAP (Rapid Extension and Active Presentation), and SPOT (Special Points On Top) with great ease.
基金supported by the National Key R&D Program of China(No.2021YFE0205700)the National Natural Science Foundation of China(Nos.62076235,62276260 and 62002356)+1 种基金sponsored by the Zhejiang Lab(No.2021KH0AB07)the Ministry of Education Industry-University Cooperative Education Program(Wei Qiao Venture Group,No.E1425201).
文摘Self-attention aggregates similar feature information to enhance the features. However, the attention covers nonface areas in face alignment, which may be disturbed in challenging cases, such as occlusions, and fails to predict landmarks. In addition, the learned feature similarity variance is not large enough in the experiment. To this end, we propose structural dependence learning based on self-attention for face alignment (SSFA). It limits the self-attention learning to the facial range and adaptively builds the significant landmark structure dependency. Compared with other state-of-the-art methods, SSFA effectively improves the performance on several standard facial landmark detection benchmarks and adapts more in challenging cases.
基金Supported by the National Natural Science Foundation of China. Acknowledgements The National Science Foundation of China supported these works. Thanks to NSFC and all the members of the research groups in Renmin University of China.
文摘Database system is the infrastructure of the modern information system. The R&D in the database system and its technologies is one of the important research topics in the field. The database R&D in China took off later but it moves along by giant steps. This report presents the achievements Renmin University of China (RUC) has made in the past 25 years and at the same time addresses some of the research projects we, RUC, are currently working on. The National Natural Science Foundation of China supports and initiates most of our research projects and these successfully conducted projects have produced fruitful results.
文摘Recommendation systems have been extensively studied over the last decade in various domains. It has been considered a powerful tool for assisting business owners in promoting sales and helping users with decision-making when given numerous choices. In this paper, we propose a novel Graph-based Context-Aware Recommendation Systems with Knowledge Graph to analyse and predict users’ behaviours, i.e., making recommendations based on historical events and their implicit associations. The model incorporates contextual information extracted from both users’ historical behaviours and events relations, where the contexts have been modelled as knowledge graphs. By leveraging the advantages offered from the knowledge graph, events dependencies and their subtle relations can be established and have been introduced in the recommendation process. Experimental results indicate that the proposed approach can outperform the state-of-the-art algorithms and achieve more accurate recommendations.
基金This work was supported by the Fundamental Research Funds for the Central Universities(2018B678X14 and 2016B44414)Postgraduate Research Practice Innovation Program of Jiangsu Province of China(KYCX18_0553 and KYLX16_0722)+1 种基金the National Natural Science Foundation of China(Grant Nos.61806137 and 61976146)Project of Natural Science Research of the Universities of Jiangsu Province(18KJB520043).
文摘Survey generation aims to generate a summary from a scientific topic based on related papers.The structure of papers deeply influences the generative process of survey,especially the relationships between sentence and sentence,paragraph and paragraph.In principle,the structure of paper can influence the quality of the summary.Therefore,we employ the structure of paper to leverage contextual information among sentences in paragraphs to generate a survey for documents.In particular,we present a neural document structure model for survey generation.We take paragraphs as units,and model sentences in paragraphs,we then employ a hierarchical model to learn structure among sentences,which can be used to select important and informative sentences to generate survey.We evaluate our model on scientific document data set.The experimental results show that our model is effective,and the generated survey is informative and readable.