In order to narrow the semantic gap existing in content-based image retrieval （CBIR）,a novel retrieval technology called auto-extended multi query examples （AMQE） is proposed.It expands the single one query image ...In order to narrow the semantic gap existing in content-based image retrieval （CBIR）,a novel retrieval technology called auto-extended multi query examples （AMQE） is proposed.It expands the single one query image used in traditional image retrieval into multi query examples so as to include more image features related with semantics.Retrieving images for each of the multi query examples and integrating the retrieval results,more relevant images can be obtained.The property of the recall-precision curve of a general retrieval algorithm and the K-means clustering method are used to realize the expansion according to the distance of image features of the initially retrieved images.The experimental results demonstrate that the AMQE technology can greatly improve the recall and precision of the original algorithms.展开更多
Multidimensional data query has been gaining much interest in database research communities in recent years, yet many of the existing studies focus mainly on ten tralized systems. A solution to querying in Peer-to-Pee...Multidimensional data query has been gaining much interest in database research communities in recent years, yet many of the existing studies focus mainly on ten tralized systems. A solution to querying in Peer-to-Peer（P2P） environment was proposed to achieve both low processing cost in terms of the number of peers accessed and search messages and balanced query loads among peers. The system is based on a balanced tree structured P2P network. By partitioning the query space intelligently, the amount of query forwarding is effectively controlled, and the number of peers involved and search messages are also limited. Dynamic load balancing can be achieved during space partitioning and query resolving. Extensive experiments confirm the effectiveness and scalability of our algorithms on P2P networks.展开更多
Dynamic programming(DP) is an effective query optimization approach to select an appropriate join order for relational database management system(RDBMS) in multi-table joins. This method was extended and made availabl...Dynamic programming(DP) is an effective query optimization approach to select an appropriate join order for relational database management system(RDBMS) in multi-table joins. This method was extended and made available in distributed DBMS(D-DBMS). The structure of this optimal solution was firstly characterized according to the distributing status of tables and data, and then the recurrence relations between a problem and its sub-problems were recursively defined. DP in D-DBMS has the same time-complexity with that in centralized DBMS, while it has the capability to solve a much more sophisticated optimal problem of multi-table join in D-DBMS. The effectiveness of this optimal strategy has been proved by experiments.展开更多
大语言模型(large language model,LLM)技术热潮对数据质量的要求提升到了一个新的高度.在现实场景中,数据通常来源不同且高度相关.但由于数据隐私安全问题,跨域异质数据往往不允许集中共享,难以被LLM高效利用.鉴于此,提出了一种LLM和...大语言模型(large language model,LLM)技术热潮对数据质量的要求提升到了一个新的高度.在现实场景中,数据通常来源不同且高度相关.但由于数据隐私安全问题,跨域异质数据往往不允许集中共享,难以被LLM高效利用.鉴于此,提出了一种LLM和知识图谱(knowledge graph,KG)协同的跨域异质数据查询框架,在LLM+KG的范式下给出跨域异质数据查询的一个治理方案.为确保LLM能够适应多场景中的跨域异质数据,首先采用适配器对跨域异质数据进行融合,并构建相应的知识图谱.为提高查询效率,引入线性知识图,并提出同源知识图抽取算法HKGE来实现知识图谱的重构,可显著提高查询性能,确保跨域异质数据治理的高效性.进而,为保证多域数据查询的高可信度,提出可信候选子图匹配算法Trust HKGM,用于检验跨域同源数据的置信度计算和可信候选子图匹配,剔除低质量节点.最后,提出基于线性知识图提示的多域数据查询算法MKLGP,实现LLM+KG范式下的高效可信跨域查询.该方法在多个真实数据集上进行了广泛实验,验证了所提方法的有效性和高效性.展开更多
随着自动驾驶技术的快速发展,准确预测周围车辆的运动轨迹成为确保行车安全的关键。现有的方法大多未充分考虑车辆与环境以及车与车之间的互动和环境情景信息,面对复杂交通场景下的轨迹预测性能不佳。基于此,提出一种融合时空查询Transf...随着自动驾驶技术的快速发展,准确预测周围车辆的运动轨迹成为确保行车安全的关键。现有的方法大多未充分考虑车辆与环境以及车与车之间的互动和环境情景信息,面对复杂交通场景下的轨迹预测性能不佳。基于此,提出一种融合时空查询Transformer的情景感知多模态车辆轨迹预测模型(Contex-aware Multimodal Vehicle Trajectory Pediction Model Based on Spatio-Temporal Query Transformer,STQformer),高效地理解和预测复杂交通环境中的车辆行为。模型以Transformer框架为基础,引入可学习的时空查询并利用社交交互模块,实现对车辆意图的深度感知和更准确的轨迹预测。实验结果表明:与当前先进的轨迹预测算法相比,STQformer在长期预测方面的性能同比性能最佳对比模型提升了9%。该模型有助于提升自动驾驶系统的安全性和可靠性,推动自动驾驶技术的发展和应用,使其更好地适应复杂多变的交通环境,减少交通事故,提高交通效率。展开更多
基金The National High Technology Research and Develop-ment Program of China (863 Program) (No.2002AA413420).
文摘In order to narrow the semantic gap existing in content-based image retrieval （CBIR）,a novel retrieval technology called auto-extended multi query examples （AMQE） is proposed.It expands the single one query image used in traditional image retrieval into multi query examples so as to include more image features related with semantics.Retrieving images for each of the multi query examples and integrating the retrieval results,more relevant images can be obtained.The property of the recall-precision curve of a general retrieval algorithm and the K-means clustering method are used to realize the expansion according to the distance of image features of the initially retrieved images.The experimental results demonstrate that the AMQE technology can greatly improve the recall and precision of the original algorithms.
基金Supported by the Natural Science Foundation ofJiangsu Province(BG2004034)
文摘Multidimensional data query has been gaining much interest in database research communities in recent years, yet many of the existing studies focus mainly on ten tralized systems. A solution to querying in Peer-to-Peer（P2P） environment was proposed to achieve both low processing cost in terms of the number of peers accessed and search messages and balanced query loads among peers. The system is based on a balanced tree structured P2P network. By partitioning the query space intelligently, the amount of query forwarding is effectively controlled, and the number of peers involved and search messages are also limited. Dynamic load balancing can be achieved during space partitioning and query resolving. Extensive experiments confirm the effectiveness and scalability of our algorithms on P2P networks.
文摘Dynamic programming(DP) is an effective query optimization approach to select an appropriate join order for relational database management system(RDBMS) in multi-table joins. This method was extended and made available in distributed DBMS(D-DBMS). The structure of this optimal solution was firstly characterized according to the distributing status of tables and data, and then the recurrence relations between a problem and its sub-problems were recursively defined. DP in D-DBMS has the same time-complexity with that in centralized DBMS, while it has the capability to solve a much more sophisticated optimal problem of multi-table join in D-DBMS. The effectiveness of this optimal strategy has been proved by experiments.
文摘大语言模型(large language model,LLM)技术热潮对数据质量的要求提升到了一个新的高度.在现实场景中,数据通常来源不同且高度相关.但由于数据隐私安全问题,跨域异质数据往往不允许集中共享,难以被LLM高效利用.鉴于此,提出了一种LLM和知识图谱(knowledge graph,KG)协同的跨域异质数据查询框架,在LLM+KG的范式下给出跨域异质数据查询的一个治理方案.为确保LLM能够适应多场景中的跨域异质数据,首先采用适配器对跨域异质数据进行融合,并构建相应的知识图谱.为提高查询效率,引入线性知识图,并提出同源知识图抽取算法HKGE来实现知识图谱的重构,可显著提高查询性能,确保跨域异质数据治理的高效性.进而,为保证多域数据查询的高可信度,提出可信候选子图匹配算法Trust HKGM,用于检验跨域同源数据的置信度计算和可信候选子图匹配,剔除低质量节点.最后,提出基于线性知识图提示的多域数据查询算法MKLGP,实现LLM+KG范式下的高效可信跨域查询.该方法在多个真实数据集上进行了广泛实验,验证了所提方法的有效性和高效性.
文摘随着自动驾驶技术的快速发展,准确预测周围车辆的运动轨迹成为确保行车安全的关键。现有的方法大多未充分考虑车辆与环境以及车与车之间的互动和环境情景信息,面对复杂交通场景下的轨迹预测性能不佳。基于此,提出一种融合时空查询Transformer的情景感知多模态车辆轨迹预测模型(Contex-aware Multimodal Vehicle Trajectory Pediction Model Based on Spatio-Temporal Query Transformer,STQformer),高效地理解和预测复杂交通环境中的车辆行为。模型以Transformer框架为基础,引入可学习的时空查询并利用社交交互模块,实现对车辆意图的深度感知和更准确的轨迹预测。实验结果表明:与当前先进的轨迹预测算法相比,STQformer在长期预测方面的性能同比性能最佳对比模型提升了9%。该模型有助于提升自动驾驶系统的安全性和可靠性,推动自动驾驶技术的发展和应用,使其更好地适应复杂多变的交通环境,减少交通事故,提高交通效率。
