Knowledge graphs(KGs)have been widely accepted as powerful tools for modeling the complex relationships between concepts and developing knowledge-based services.In recent years,researchers in the field of power system...Knowledge graphs(KGs)have been widely accepted as powerful tools for modeling the complex relationships between concepts and developing knowledge-based services.In recent years,researchers in the field of power systems have explored KGs to develop intelligent dispatching systems for increasingly large power grids.With multiple power grid dispatching knowledge graphs(PDKGs)constructed by different agencies,the knowledge fusion of different PDKGs is useful for providing more accurate decision supports.To achieve this,entity alignment that aims at connecting different KGs by identifying equivalent entities is a critical step.Existing entity alignment methods cannot integrate useful structural,attribute,and relational information while calculating entities’similarities and are prone to making many-to-one alignments,thus can hardly achieve the best performance.To address these issues,this paper proposes a collective entity alignment model that integrates three kinds of available information and makes collective counterpart assignments.This model proposes a novel knowledge graph attention network(KGAT)to learn the embeddings of entities and relations explicitly and calculates entities’similarities by adaptively incorporating the structural,attribute,and relational similarities.Then,we formulate the counterpart assignment task as an integer programming(IP)problem to obtain one-to-one alignments.We not only conduct experiments on a pair of PDKGs but also evaluate o ur model on three commonly used cross-lingual KGs.Experimental comparisons indicate that our model outperforms other methods and provides an effective tool for the knowledge fusion of PDKGs.展开更多
Metallic nanoparticle （NP） shapes have a significant influence on the property of composite embedded with metallic NPs. Swift heavy ion irradiation is an effective way to modify shapes of metallic NPs embedded in an...Metallic nanoparticle （NP） shapes have a significant influence on the property of composite embedded with metallic NPs. Swift heavy ion irradiation is an effective way to modify shapes of metallic NPs embedded in an amorphous matrix. We investigate the shape deformation of Ag NPs with irradiation fluence, and 357 MeV Ni ions are used to irradiate the silica containing Ag NPs, which are prepared by ion implantation and vacuum annealing. The UV-vis results show that the surface plasmon resonance （SPR） peak from Ag NPs shifts from 400 to 377nm. The SPR peak has a significant shift at fluence lower than 1 × 10^14 ions/cm2 and shows less shift at fluence higher than 1 × 10^14 ions/cm2. The TEM results reveal that the shapes of Ag NPs also show significant deformation at fluence lower than 1 × 10^14 ions/cm2 and show less deformation at fluence higher than 1 × 10^14 ions/cm2. The blue shift of the SPR peak is considered to be the consequence of defect production and Ag NP shape deformation, Based on the thermal spike model calculation, the temperature of the silica surrounding Ag particles first increases rapidly, then the region of Ag NPs close to the interface of Ag/silica is gradually heated. Therefore, the driven force of Ag NPs deformation is considered as the volume expansion of the first heated silica layer surrounding Ag NPs.展开更多
The means of knowledge graph embedding is to transform entities and relations into low-dimensional vectors.When it is necessary to obtain the embedding results of two hetero-field knowledge graphs in a unified vector ...The means of knowledge graph embedding is to transform entities and relations into low-dimensional vectors.When it is necessary to obtain the embedding results of two hetero-field knowledge graphs in a unified vector space,there are only a few aligned entities between them,previous methods first need to merge the two graphs into a large graph,and then re-embed the entire large graph.This ignores the potential reuse of the original representation embeddings of two knowledge graphs and will lead to a lot of time consumption.To address this problem,this paper proposes a hetero-field knowledge graph embedding fusion model(BlockEF)based on aligned entities.According to the fact that the aligned entities of the two graphs should be located in the same position in the vector space,the transformation relationship between the embeddings of two graphs is firstly obtained,and then graph embedding is fine-tuned and optimized to achieve efficient fusion of hetero-field knowledge graph embeddings.The experimental results show that our method can significantly reduce the computational burden of heterofield knowledge graph embedding fusion and ensure the quality of embedding fusion.展开更多
融合子图学习与联邦学习后,联邦子图学习在保护数据隐私的同时可实现多客户端子图信息之间的协同学习.然而,由于不同客户端的数据收集方式存在差异,图数据通常呈现非独立同分布特性.同时,不同客户端局部图数据的结构和特征也存在较大差...融合子图学习与联邦学习后,联邦子图学习在保护数据隐私的同时可实现多客户端子图信息之间的协同学习.然而,由于不同客户端的数据收集方式存在差异,图数据通常呈现非独立同分布特性.同时,不同客户端局部图数据的结构和特征也存在较大差异.这些因素导致联邦子图学习在训练过程中出现收敛困难和泛化能较差等问题.为了解决此问题,文中提出基于嵌入对齐与参数激活的个性化联邦子图学习方法(Personalized Federated Subgraph Learning with Embedding Alignment and Parameter Activation,FSL-EAPA).首先,根据客户端之间的相似性进行个性化模型聚合,降低数据非独立同分布对整体性能的影响.然后,引入参数选择性激活进行模型更新,应对子图结构特征的异质性.最后,利用更新后的客户端为各本地节点嵌入提供正负聚类表示,聚集同类局部节点.因此,FSL-EAPA能充分学习各节点的特征表示,较好地适应不同客户端之间的差异化数据分布.在真实基准图数据集上的实验表明FSL-EAPA的有效性,并且在不同场景下都能获得较高的分类精度.展开更多
基金supported by the National Key R&D Program of China(2018AAA0101502)the Science and Technology Project of SGCC(State Grid Corporation of China):Fundamental Theory of Human-in-the-Loop Hybrid-Augmented Intelligence for Power Grid Dispatch and Control。
文摘Knowledge graphs(KGs)have been widely accepted as powerful tools for modeling the complex relationships between concepts and developing knowledge-based services.In recent years,researchers in the field of power systems have explored KGs to develop intelligent dispatching systems for increasingly large power grids.With multiple power grid dispatching knowledge graphs(PDKGs)constructed by different agencies,the knowledge fusion of different PDKGs is useful for providing more accurate decision supports.To achieve this,entity alignment that aims at connecting different KGs by identifying equivalent entities is a critical step.Existing entity alignment methods cannot integrate useful structural,attribute,and relational information while calculating entities’similarities and are prone to making many-to-one alignments,thus can hardly achieve the best performance.To address these issues,this paper proposes a collective entity alignment model that integrates three kinds of available information and makes collective counterpart assignments.This model proposes a novel knowledge graph attention network(KGAT)to learn the embeddings of entities and relations explicitly and calculates entities’similarities by adaptively incorporating the structural,attribute,and relational similarities.Then,we formulate the counterpart assignment task as an integer programming(IP)problem to obtain one-to-one alignments.We not only conduct experiments on a pair of PDKGs but also evaluate o ur model on three commonly used cross-lingual KGs.Experimental comparisons indicate that our model outperforms other methods and provides an effective tool for the knowledge fusion of PDKGs.
基金Supported by the National Natural Science Foundation of China under Grant Nos 11475230 and U1532262
文摘Metallic nanoparticle （NP） shapes have a significant influence on the property of composite embedded with metallic NPs. Swift heavy ion irradiation is an effective way to modify shapes of metallic NPs embedded in an amorphous matrix. We investigate the shape deformation of Ag NPs with irradiation fluence, and 357 MeV Ni ions are used to irradiate the silica containing Ag NPs, which are prepared by ion implantation and vacuum annealing. The UV-vis results show that the surface plasmon resonance （SPR） peak from Ag NPs shifts from 400 to 377nm. The SPR peak has a significant shift at fluence lower than 1 × 10^14 ions/cm2 and shows less shift at fluence higher than 1 × 10^14 ions/cm2. The TEM results reveal that the shapes of Ag NPs also show significant deformation at fluence lower than 1 × 10^14 ions/cm2 and show less deformation at fluence higher than 1 × 10^14 ions/cm2. The blue shift of the SPR peak is considered to be the consequence of defect production and Ag NP shape deformation, Based on the thermal spike model calculation, the temperature of the silica surrounding Ag particles first increases rapidly, then the region of Ag NPs close to the interface of Ag/silica is gradually heated. Therefore, the driven force of Ag NPs deformation is considered as the volume expansion of the first heated silica layer surrounding Ag NPs.
基金supported by the Deep-time Digital Earth (DDE) Big Science Programin part by the National Natural Science Foundation of China (NSFC) (No. 61972365)in part by the Open Research Project of the Hubei Key Laboratory of Intelligent Geo-Information Processing (No. KLIGIP-2021B01)
文摘The means of knowledge graph embedding is to transform entities and relations into low-dimensional vectors.When it is necessary to obtain the embedding results of two hetero-field knowledge graphs in a unified vector space,there are only a few aligned entities between them,previous methods first need to merge the two graphs into a large graph,and then re-embed the entire large graph.This ignores the potential reuse of the original representation embeddings of two knowledge graphs and will lead to a lot of time consumption.To address this problem,this paper proposes a hetero-field knowledge graph embedding fusion model(BlockEF)based on aligned entities.According to the fact that the aligned entities of the two graphs should be located in the same position in the vector space,the transformation relationship between the embeddings of two graphs is firstly obtained,and then graph embedding is fine-tuned and optimized to achieve efficient fusion of hetero-field knowledge graph embeddings.The experimental results show that our method can significantly reduce the computational burden of heterofield knowledge graph embedding fusion and ensure the quality of embedding fusion.
文摘融合子图学习与联邦学习后,联邦子图学习在保护数据隐私的同时可实现多客户端子图信息之间的协同学习.然而,由于不同客户端的数据收集方式存在差异,图数据通常呈现非独立同分布特性.同时,不同客户端局部图数据的结构和特征也存在较大差异.这些因素导致联邦子图学习在训练过程中出现收敛困难和泛化能较差等问题.为了解决此问题,文中提出基于嵌入对齐与参数激活的个性化联邦子图学习方法(Personalized Federated Subgraph Learning with Embedding Alignment and Parameter Activation,FSL-EAPA).首先,根据客户端之间的相似性进行个性化模型聚合,降低数据非独立同分布对整体性能的影响.然后,引入参数选择性激活进行模型更新,应对子图结构特征的异质性.最后,利用更新后的客户端为各本地节点嵌入提供正负聚类表示,聚集同类局部节点.因此,FSL-EAPA能充分学习各节点的特征表示,较好地适应不同客户端之间的差异化数据分布.在真实基准图数据集上的实验表明FSL-EAPA的有效性,并且在不同场景下都能获得较高的分类精度.
