17 results about "Structure relation" patented technology

In implementations of systems for generating templates using structure-based matching, a computing device implements a template system to receive input data describing a set of digital design elements. The template system represents the input data as a sentence in a design structure language that describes structural relationships between design elements included in the set of digital design elements. An input template embedding is generated based on the sentence in the design structure language. The template system generates a digital template that includes the set of digital design elements for display in a user interface based on the input template embedding.

The invention discloses a building quality hidden danger analysis method and system based on cloud computing software, relates to the field of building quality hidden danger analysis based on cloud computing software, is used for enhancing the early diagnosis and early warning capability of systematic and global major hidden dangers of a building structure, and comprises the following steps: obtaining the position of a defect; determining a first building component corresponding to the position of the defect in the building information model according to the position of the defect; verifying the first building component based on a pre-defined inter-component structural relationship rule set in the building information model; the inter-component structural relationship rule set is used for indicating structural relationships among different building components; associating the first building component with the defect under the condition that the verification of the first building component is successful; and under the condition that the verification of the first building component fails, determining a second building component corresponding to the defect according to the structural relationship rule set between the components, and associating the second building component with the defect.

This application provides a representation learning method and system for a unified structural state. The representation learning method includes acquiring multi-source heterogeneous data of the same lipid nanoparticle object; organizing the multi-source heterogeneous data into object-level associations using object identifiers; encoding the multi-source heterogeneous data separately and mapping them to a unified latent space to obtain an initial representation of the lipid nanoparticle object; extracting prior information characterizing the internal structural relationships and / or thermodynamic states of the particles; constructing training constraint signals based on the prior information; using the training constraint signals as constraint targets to constrain the representation relationships within the unified latent space; jointly training the model using the basic task loss and the constraint loss corresponding to the training constraint signals; and using the unified structural state representation as a shared backbone representation for reuse in at least one downstream task. This improves the physical consistency, transferability, and engineering applicability of the lipid nanoparticle representation learning results.

The invention discloses an electrical secondary drawing component detection method and device based on a structure perception mechanism, and a medium, and belongs to the technical field of drawing component detection, and the method comprises the steps: collecting an electrical secondary drawing, carrying out the unified rasterization and preprocessing of the electrical secondary drawing, and obtaining a preprocessed electrical secondary drawing; constructing an electrical secondary drawing component detection model, inputting the preprocessed electrical secondary drawing, and outputting the category probability, confidence and bounding box parameters of each target; in a training stage of an electrical secondary drawing component detection model, geometric centers of all symbols on an electrical secondary drawing are taken as nodes, a symbol structure relation graph is constructed according to alignment, adjacency and electrical connection rules, and a loss function is constructed. According to the invention, the problems of false detection and missing detection are significantly reduced, and high-precision identification is realized; the method has high detection efficiency, adapts to complex scenes such as aged drawings and dense layouts, has smaller model size and higher deployment efficiency, and is low in cost and high in practicability.

According to the code structured management system and method provided by the embodiment of the invention, a code splitting module performs structured splitting on a source code file in a local code library to generate code snippets with independent semantics; the context determination module determines the context of each code snippet, wherein the context comprises a code context, a calling context or a semantic context; the structure storage module constructs a corresponding structure relation graph based on the context of the code snippets, and stores the structure relation graph in a local storage space; and the interaction module responds to a viewing instruction for any code snippet, extracts the current code snippet and the corresponding context based on the structural relation graph, and displays the code snippet and the context. By splitting the source code into the code snippets and constructing the structure relation graph based on the contexts of the code snippets, the logic relation between the code snippets can be visually displayed, and the understanding speed and management efficiency of developers on the source code are improved.

The invention relates to the technical field of internet of things intrusion detection, in particular to a double-queue memory and structural relationship distillation method for intrusion detection, which comprises the following steps of: enabling a model to learn feature representation with distinction degree through a double-queue collaborative memory network under the condition of completely not needing to mark data; according to the method, distance and angle relations in a sample feature space are comprehensively modeled by using a symmetric spatial structure relation distillation method, and a student model of lightweight design efficiently learns structural relation knowledge in the feature space of a teacher network through a teacher-student distribution bidirectional alignment mechanism, and linear fine adjustment is performed on the student model through a small number of labels. The effective intrusion detection system which is lightweight, can process data in real time and can run in an unsupervised mode is obtained, the problem of label dependence is relieved, and the method is suitable for an Internet of Things scene with limited resources.

The invention provides a multi-feature fusion entity alignment method for a standard text, and relates to the technical field of standardized professional knowledge graph construction, and the method comprises the following steps: S1, determining an entity type: determining the entity type needing to be aligned in a standard entity; s2, performing preliminary screening of aligned entities by using a term library: performing preliminary screening of entities in a standard by using the term library, and determining an entity range in which alignment work needs to be performed by traversing all entities in the knowledge graph; s3, carrying out attribute similarity judgment; s4, carrying out structural relationship similarity judgment; and S5, performing entity fusion and updating. According to the method, the entity alignment method is constructed from three dimensions of entity name expression, entity attribute and entity structural relationship, ambiguous entities such as'synonym and different names' and'same name and different meanings' in the standard can be comprehensively covered, and the problem of entity redundancy caused by name difference or expression difference is more quickly and accurately solved through a multi-time and multi-dimensional screening mode.

The invention provides an ultra-large BOM design method based on streaming constraint propagation and thinking chain reasoning cascading. The method comprises the steps that a target function requirement set is acquired; analyzing the target function demand set, and determining a BOM design initial state through a preset function-structure mapping rule in combination with a historical structure relationship set; executing thinking chain reasoning in the BOM design initial state to generate a candidate design state; performing engineering rule judgment and path cutting on the candidate design states based on streaming constraint propagation, and outputting consistency confirmation states meeting conditions; wherein the engineering rule judgment adopts an execution mode of an engineering rule engine; structural convergence is performed based on the consistency acknowledgement state and a final standardized BOM is generated. According to the method, the engineering controllability and delivery efficiency of BOM design of large equipment are improved.

The application discloses a kind of bird image recognition analysis methods based on computer vision, comprising the following steps: obtaining input image containing bird target, generating structure node set and structure relationship diagram;Analysis structure node combination relationship, generate structure lock candidate set;To candidate structure mode distribution structure lock credibility, generate structure lock credibility sequence;Structure energy propagation relationship is built, and structure energy propagation result is generated;Update structure node relationship state, generate visual structure evolution sequence;Calculate structure node energy distribution and judge whether into preset structure stable domain;Identify the node combination that leads to structure instability and update structure lock candidate set;When candidate structure mode enters structure stable domain, the corresponding bird recognition analysis result is output.The application realizes bird image recognition by structure energy propagation and visual structure evolution analysis, with the advantages of high recognition stability and strong analysis ability.

The invention discloses an equipment knowledge graph cross-language instance alignment method, and relates to the technical field of equipment knowledge graph technologies, which comprises the following specific steps of: carrying out embedded learning on structure and relation characteristics in a knowledge graph by utilizing a GCN and combining a Highway mechanism in a structure and relation joint embedding module, respectively generating structure embedding and relation embedding, and carrying out structure embedding and relation embedding on the structure and relation characteristics in a structure and relation joint embedding module; the two are fused into structural relation joint embedding through joint learning, and attribute information of instances in the knowledge graph is modeled through GCN in an attribute embedding module to generate attribute embedding; in a semi-supervised seed set enhancement module, a seed set is expanded through a semi-supervised bidirectional nearest neighbor iteration strategy, the problem of training data scarcity is relieved, and the quality of the seed set is optimized; and the structural relation joint embedding and the attribute embedding are fused into final embedding in the alignment module. According to the invention, the structure relation joint embedding module, the attribute embedding module, the semi-supervised seed set enhancement module and the alignment module are arranged, so that the alignment precision is improved.

The application discloses a multilevel system model construction method, system, device and medium, and relates to the technical field of system engineering and system modeling. The method comprises the following steps: acquiring information data of a multilevel system to be constructed; determining a structure undirected graph according to the structure relationship and the construction objects between construction objects; performing regional segmentation according to a set path step based on the level corresponding to the construction objects and the structure undirected graph, and obtaining a plurality of regional network structures; for any regional network structure, determining the dependency relationship between the construction objects by using a data structure learning algorithm, and determining a graphical network according to the dependency relationship; determining a structure network model of the multilevel system to be constructed according to the structure undirected graph and all the graphical networks; and the structure network model is used for representing the mechanism characteristic information of the multilevel system in cooperative interaction. The application can perfect the global structure of the system and realize the construction of a global static model.

In implementations of systems for generating templates using structure-based matching, a computing device implements a template system to receive input data describing a set of digital design elements. The template system represents the input data as a sentence in a design structure language that describes structural relationships between design elements included in the set of digital design elements. An input template embedding is generated based on the sentence in the design structure language. The template system generates a digital template that includes the set of digital design elements for display in a user interface based on the input template embedding.

The application discloses a kind of dam safety monitoring space-time knowledge graph automatic construction and evolution method, it is related to dam safety monitoring technical field, including: the three-dimensional coordinate data of acquisition monitoring point and the data of dam body structure model, the spatial adjacent relationship and spatial inclusion relation between monitoring point are automatically identified, and spatial relationship set is established;Combined with dam body structure model, the structural relationship set between monitoring point and dam body structure unit is automatically established;The time series data of monitoring point is acquired and preprocessed, and time series relationship set is established;Space-time knowledge graph is constructed;Real-time monitoring dam change event, dynamically evolving space-time knowledge graph, consistency verification and rollback are carried out to updated relationship, dam safety state reasoning and early warning are carried out, and change audit log is recorded.The application realizes the full-automatic construction and dynamic updating of dam safety monitoring space-time knowledge graph, provides efficient, accurate, dynamic knowledge graph construction method for dam safety monitoring.

The application provides a representation learning method and system for unified structure state. The representation learning method comprises: obtaining multi-source heterogeneous data of the same lipid nanoparticle object, and performing object-level association organization on the multi-source heterogeneous data by using object identification; mapping the multi-source heterogeneous data to a unified hidden space after encoding to obtain initial representation of the lipid nanoparticle object; extracting prior information representing internal structure relationship and / or thermodynamic state of the particle; constructing a training constraint signal based on the prior information, and constraining the representation relationship in the unified hidden space by taking the training constraint signal as a constraint target; performing joint training by a joint training model through a basic task loss and a constraint loss corresponding to the training constraint signal; and taking the unified structure state representation as a shared backbone representation for reuse by at least one downstream task. In this way, the physical consistency, transferability and engineering applicability of the representation learning result of the lipid nanoparticle can be improved.

The application proposes a super-large BOM design method based on flow constraint propagation and thinking chain reasoning cascade, which comprises the following steps: obtaining a target function requirement set; analyzing the target function requirement set, combining a historical structure relationship set, and determining a BOM design initial state through a preset function-structure mapping rule; executing thinking chain reasoning on the BOM design initial state to generate a candidate design state; performing engineering rule judgment and path pruning on the candidate design state based on flow constraint propagation to output a consistent confirmation state that meets the conditions; wherein the engineering rule judgment adopts the execution mode of an engineering rule engine; and performing structure convergence based on the consistent confirmation state and generating a final standardized BOM. The application improves the engineering controllability and delivery efficiency of large equipment BOM design.

This invention discloses an online handwritten mathematical formula generation method based on a structure-aware latent space diffusion model. First, the input LaTeX expression is parsed to obtain symbol content, spatial topological relationships between symbols, and hierarchical depth information of the symbols, and a structured representation (RelAST) is constructed. Second, a generation framework based on a latent space conditional diffusion model is constructed, and a structure-aware sequence variational autoencoder is designed. By introducing symbol recognition constraints and structural relationship constraints, the latent representation simultaneously represents symbol semantics and spatial topological structure information. In the generation stage, using the structured representation (RelAST) as conditional input and combining prior information on the number of symbols, the target latent variable representation is generated in the latent space through the diffusion model. Finally, the decoder recovers the complete online handwritten trajectory sequence. This invention can effectively model the two-dimensional structural relationships of mathematical formulas, achieving unified generation of symbol content and spatial layout.

This invention relates to the field of interface management technology, specifically to a method for managing interfaces inside and outside the boundary of a target object. The method includes the following steps: identifying the target object and associated objects assembled with it; structurally disassembling the target object and associated objects to obtain the internal structural components of the target object and the external structural components of the associated objects; setting interfaces representing these structural relationships based on the structural relationships; constructing an interface matrix based on the structural relationships; and transmitting interface information to synchronize the interface information. This invention, by structurally disassembling the target object and associated objects, analyzes the structural relationships between the disassembled components, sets interfaces to represent these structural relationships for the components, constructs an interface matrix based on the structural relationships, manages the component interfaces, and transmits interface information between the target object and associated objects at the component production operation end, maintaining information interaction and ensuring the functionality and assembly compatibility between components.