89 results about "Graph spectra" patented technology

This invention provides a method and system for constructing a cross-device, cross-modal user memory graph. The method includes acquiring multimodal data from multiple devices and applications, uniformly parsing it into standardized memory atomic events with timestamps, calculating causal relationships between adjacent events during real-time processing, linking highly correlated event pairs into instantaneous event chains, performing deep encoding of the event chains using an offline dual-stream temporal semantic embedding model, and inferring candidate graph substructures using a graph neural network. When fusing these substructures into the user's main graph, relationship conflicts are resolved through confidence comparison, retaining highly reliable results, identifying stable subgraphs, and using a multi-head attention graph summarization network to abstract the subgraphs into higher-order memory concept nodes, thereby achieving continuous optimization and updating of the graph.

This invention provides a user profiling and adaptive transaction permission control method for duty-free e-commerce. It belongs to the fields of big data processing and artificial intelligence technology. The invention includes: collecting user status and operation data, generating a fusion feature matrix through feature fusion; injecting this matrix into heterogeneous graph nodes to construct a dynamic graph, and mining hidden group topological features through a graph attention mechanism; inputting the topological features into a large model to deeply mine the intent of proxy purchasing, and outputting a comprehensive user risk confidence score; constructing an environment-aware threshold adjustment center to dynamically adjust risk control thresholds; matching the comprehensive risk confidence score with the dynamic thresholds to execute tiered permission control. This invention achieves the synergy of big data and a large model, enabling accurate identification of hidden proxy purchasing, reducing false positives while ensuring risk control, balancing security and revenue, and realizing user risk identification and permission control in duty-free e-commerce.

The application relates to the technical field of transformer partial discharge identification, and discloses a transformer partial discharge identification method and device. The method comprises the following steps: acquiring multi-terminal high-frequency pulse current signals through a multi-terminal physical sensing array arranged on a transformer electric circuit; extracting time-domain features, frequency-domain features and polarity features from the multi-terminal high-frequency pulse current signals, and constructing a time-frequency-polarity three-dimensional tensor atlas; constructing a topological graph model of the transformer based on the physical structure of the transformer and a measured injection test; inputting the time-frequency-polarity feature tensor atlas and the topological graph model into a graph neural network, performing feature aggregation on the time-frequency-polarity feature tensor atlas based on the topological graph model; performing global graph feature pooling and probability classification on the aggregated features output by the graph neural network, and obtaining an identification result of transformer partial discharge. The application can solve the problem of high misjudgment rate of transformer partial discharge detection in a complex electromagnetic environment.

The application discloses a long-term service hydroelectric generating set operation state evaluation and early warning method and system, and belongs to the technical field of hydroelectric generating set state monitoring and fault early warning. The application converts a traditional point-to-point regression problem into a signal recovery problem in graph signal processing based on a graph smoothing regression method, integrates and constructs a high-dimensional working condition space adaptive neighbor graph by integrating a training set and a to-be-measured set, calculates an edge weight through an adaptive kernel bandwidth and constructs a graph Laplace matrix, introduces a variational objective function with global potential energy anchoring to complete full graph signal analytical solution, obtains a health benchmark value corresponding to a working condition, and realizes unit state evaluation and adaptive early warning through residual comparison of a measured value and the health benchmark value. The application has strong robustness, high full-working-condition evaluation accuracy, and can stably anchor the health benchmark without fault samples, accurately identifies early performance degradation of the unit, and is suitable for state maintenance application of long-term service pumped storage units.

The application provides a material similarity analysis method based on a semantic library and a knowledge graph, comprising the following steps: S1: judging whether the decisive attributes of two materials are similar, if any of the decisive attributes is not similar, then the two materials are non-redundant materials, otherwise, entering S2; S2: calculating the similarity of the field part according to the non-decisive attribute weight and the distance of each field; S3: combining all the knowledge graph paths and weight values to calculate the similarity of the graph part; S4: adding the similarity of the field part and the similarity of the graph part to obtain the final material similarity, and judging whether the two materials are high similarity through the final material similarity. The material similarity analysis method provided by the application optimizes the accuracy of the similarity algorithm.

This invention provides a method, system, and device for temporal knowledge graph reasoning based on graph topology enhancement, which can be applied to the field of artificial intelligence technology. The method includes: adding nodes and directed edges to a historical topology graph based on the types of multiple elements in the quadruple to be completed, to obtain a target topology graph; determining the potential association strength between the quadruple to be completed and multiple historical quadruples based on the time decay of each historical quadruple and the connection between each node and each historical node in the target topology graph; constructing prompt words based on the quadruple to be completed and at least one target quadruple determined using the potential association strength; inputting the prompt words into a large language model to infer multiple first candidate elements and the first confidence level of each first candidate element; and determining the element to be completed based on the first candidate elements, the first confidence levels of the first candidate elements, and the second candidate elements and their second confidence levels obtained from the graph model reasoning.

The present application belongs to but is not limited to the technical field of shale gas production, and particularly relates to a method and system for ontology graph construction and abnormal cause-effect chain tracing, comprising: S1, well, pry, process, pressure difference, alarm, operation event object and relationship mode; S2, cause-effect chain coding: forming a calculable path of 'upstream equipment-process-downstream abnormality'; S3, realizing bidirectional linkage of ontology graph and detection result. The ontology knowledge graph of different business scenarios in the field of shale gas is constructed, and the well, pry, process, sensor entity and their relationships are defined; when the abnormal detection model outputs the result, the result is mapped to the graph; through graph query or graph traversal algorithm, the upstream and downstream associated entities of the abnormal node are automatically traced, and natural language explanation is generated; an interface supporting semantic query is provided, and the user is allowed to ask questions in an exploratory manner.

The application discloses a substation work ticket knowledge graph construction method based on a large language model, and belongs to the field of energy power and artificial intelligence. The method comprises the following steps: based on standardized electrical drawing annotation data, independent knowledge graphs are generated through data reorganization preprocessing, rough knowledge graph construction and attribute supplement reasoning; all independent knowledge graphs are divided into framework layer graphs and detail layer graphs according to voltage levels, and hierarchical fusion strategies are adopted to complete full graph fusion; breakpoint detection is carried out by using an isolated point detection algorithm based on degree centrality, connectivity analysis is carried out through depth-first search, and non-connected subgraphs are re-fused; abnormal nodes are screened based on electrical connection rules, a large language model is input to automatically fill in Agent to generate correction suggestions, and graph optimization is completed. Through the cooperation of the large model Agent and multiple algorithms, the application improves the efficiency and accuracy of electrical data structuring, and provides knowledge network support for substation automation operation and fault diagnosis.

The application relates to the technical field of multi-modal information processing and intelligent health management, and discloses a multi-modal AI data tracing and early warning system and a portable sampling device thereof.The system comprises a multi-modal data preprocessing and feature engineering module, which is used for fusing spectrum, geographical space, image and text data into a unified high-dimensional feature vector; a dynamic space-time pollution atlas construction module, which is used for establishing and dynamically maintaining a nodal space-time graph structure; and a tracing analysis module based on a graph neural network, which is used for calculating a contribution score on the graph structure to determine the most important space-time source.The portable multi-modal sampling device is provided with a shell with a man-machine interaction interface, and a sample pretreatment module integrated with an integrated micro-flow channel and a replaceable unit is integrated inside.The application solves the problems of multi-source heterogeneous data fusion difficulty, poor real-time tracing and lack of a prediction mechanism, improves the real-time tracing analysis and increases the prediction ability of future risks.

The application discloses an industrial chain analysis method and related products. The industrial chain analysis method comprises the following steps: obtaining a plurality of initial upstream graph sub-trees, initial midstream graph sub-trees and initial downstream graph sub-trees of a target industrial chain according to related knowledge of the target industrial chain; merging the plate level and / or the link level in the plurality of initial graph sub-trees of the same type to obtain a merged graph sub-tree; and determining the corresponding enterprises and the head enterprises of each link of the link level in each merged graph sub-tree to analyze the target industrial chain. In the embodiment of the application, the plate level and the link level in different initial graph sub-trees are merged in the process of constructing the industrial chain, so that the granularity of the industrial chain graph is appropriate, and the macro graph can be grasped.

This application relates to a method and system for fault monitoring of thermal flow sensors based on multi-source data fusion. The method includes: dividing historical operating status data at various time points into multiple data subsets according to a preset time window; updating a first graph object based on the features of each graph node and the correlation strength of edge features corresponding to the data subsets to construct a time-series subgraph corresponding to multiple fault category labels and training a first defect detection model; dividing current operating status data into a time-series subgraph to be detected corresponding to multiple data subsets to be detected according to a preset time window; inputting the time-series subgraph to be detected into the first defect detection model to determine the preliminary fault detection result of the thermal flow sensor to be detected; and correcting the first fault category based on domain knowledge graph and text feature data to determine a second fault category and its corresponding natural language description. This application can accurately identify complex faults caused by multi-parameter collaboration.

The present application relates to the technical field of electron microscope image analysis, and particularly relates to a transmission electron microscope micro-area diffraction crystal direction analysis method combined with a graph neural network, comprising: pre-processing a diffraction image, extracting a spot and converting it into a reciprocal space vector node; verifying the sum of any three node vectors and a deviation, establishing a connection edge if the deviation is within a threshold value, and constructing a constraint topological graph; inputting the constraint topological graph into a graph neural network, calculating edge weights and aggregating information according to neighbor node strength and reciprocal space vectors, and outputting each node crystal plane index prediction; reconstructing a theoretical reciprocal space vector based on the prediction result, calculating an overall error, and if the error is out of limit, iteratively adjusting parameters until a requirement is met or a maximum number of times is reached, and outputting a final result, and outputting a final crystal direction analysis result. The present application introduces crystallography priori through a constraint topological graph, combines with a graph neural network iterative optimization, and improves the crystal direction analysis accuracy and robustness of a complex diffraction pattern.

The application discloses a retrieval method and a retrieval system. The retrieval method is applied to heterogeneous knowledge data, and the heterogeneous knowledge data includes index type structure knowledge, narrative type structure knowledge and graph type structure knowledge. The retrieval method comprises the following steps: acquiring the heterogeneous knowledge data, constructing an adaptive graph structure for the heterogeneous knowledge data, generating a graph structure representation signal according to the graph structure, wherein a multi-level index graph is constructed for the index type structure knowledge, a keyword index graph is constructed for the narrative type structure knowledge, and a knowledge graph is constructed for the graph type structure knowledge; receiving a retrieval question and the graph structure representation signal, performing retrieval reasoning based on a preset retrieval model, obtaining an evidence chain supporting the retrieval question, outputting a retrieval answer according to the evidence chain, and generating evidence traceability data. The technical scheme of the application can effectively improve the accuracy of the retrieval result, improve the retrieval efficiency, and meet the efficient utilization of the heterogeneous knowledge data.

The application relates to the technical field of knowledge graph, and particularly provides a strabismus three-dimensional mapping method and system based on a knowledge graph, which comprises the following steps: taking a starting point space coordinate, a patient corneal reflection point offset and an eyeball rotation center coordinate as attribute values of an entity node, establishing a directed connection edge between nodes according to a fixed attachment relationship between muscles and eyeballs and between muscles and bone walls in anatomy, and constructing an initial heterogeneous graph by all entity nodes and the directed connection edge; weighting and aggregating attribute values of neighborhood nodes connected by the directed connection edge of each node in the obtained initial heterogeneous graph to obtain an aggregated feature vector of the node; and mapping all node coordinates in each obtained sub-graph to a unified three-dimensional coordinate system, and adjusting the relative positions between the sub-graphs so that all positioning base points are located on a plane. The application realizes systematic mapping and pathological posture analysis of three-dimensional space configurations of eye external muscles of a strabismus patient.

The present application relates to the technical field of knowledge reasoning and graph structure information retrieval, and particularly relates to a travel multi-source heterogeneous data fusion and knowledge graph construction method. The method comprises: obtaining travel multi-source heterogeneous data, and generating pretreatment data through differentiation processing; combining a travel field knowledge base and a relationship template library for structured processing to obtain standardized records; extracting target entities and constructing a keyword matrix to generate candidate triplets under the constraint of the template library; obtaining target triplets through confidence calculation and screening, and then performing entity alignment and fusion; performing explicit and implicit double conflict detection and knowledge logic verification on the fused triplets to obtain conflict-free graph content; incrementally writing the graph into a Neo4j graph and configuring version backtracking and error correction entry for travel recommendation, route planning and user error correction feedback retriggering. The present application can effectively improve the data quality and reliability of travel knowledge graph construction, and support version backtracking and user collaborative error correction.

The application discloses a geometric embedding knowledge graph question answering method based on a large language model, first, aiming at the structural heterogeneity of knowledge graph chain, level and cycle, entity and relationship embedding is learned in Euclidean space, hyperbolic space (Poincare ball model) and spherical space (unit sphere) respectively, and the triple score of the three geometric spaces is optimized by combining negative sampling and maximum interval loss. Secondly, in the reasoning stage, the large language model analyzes the problem and identifies the key entity, constructs the relevant subgraph, and generates the candidate path by using the breadth-first search or beam search; the path node is projected in three ways and spliced to calculate the attention weight, the fusion node representation and edge level fusion score are obtained, and the path score is accumulated. Finally, the structured prompt is formed by selecting the Top-K evidence path and its weight, contribution value, the large model is constrained to generate the answer and explanation, the traceability and accuracy are improved, the consistency is enhanced, and the illusion is reduced.

This invention discloses a cross-graph entity alignment method driven by controllable graph attention and candidate semantics, comprising the following steps: Step 1, construction of multi-view controllable entity representation; Step 2, generation of candidate entities with high recall; Step 3, selection of candidate set semantics; Step 4, confidence calibration and rejection. In the candidate generation stage, this invention effectively suppresses the propagation and accumulation of heterogeneous structural noise through multi-view controllable entity representation and a dual-path controllable graph attention mechanism, significantly improving the discriminativeness and scene adaptability of entity representation. By introducing a Soft-InfoNCE soft contrast learning objective in the candidate recall optimization process, the recall rate of truly matching entities is significantly improved. In the fine-ranking stage, a candidate set joint semantic discrimination mechanism overcomes the limitations of independent matching for each candidate, reducing the risk of mismatch. A dual-threshold rejection mechanism effectively identifies suspended entities, avoiding misalignment problems in open scenes.

This invention provides a method and system for abnormal event association processing based on an operation and maintenance knowledge graph, belonging to the field of operation and maintenance management technology. The method includes: taking each abnormal event entity in a structured abnormal event sequence as an input node set; based on the topological mapping position of the input node set in a preset operation and maintenance knowledge graph, aggregating the k-order adjacency topological features and business semantic embedding features of each input node to obtain a graph event representation vector; performing pairwise comparison and temporal alignment of the feature vectors representing different abnormal event entities in the graph event representation vector; iteratively calculating the temporal lag correlation index and state transition probability matrix between each abnormal event entity; and converging to obtain a candidate abnormal event association subgraph based on the state transition probability matrix and a preset association confidence threshold. This invention realizes abnormal event association and handling instruction generation, improving operation and maintenance efficiency and the accuracy of fault root cause location.

The application discloses a photovoltaic power generation typical scene extraction method based on multi-dimensional similarity and adaptive graph fusion, and belongs to the technical field of new energy consumption and random optimization of power systems. The method firstly carries out standardization and time aggregation pretreatment on historical photovoltaic output data of multiple power stations; secondly, a comprehensive similarity (PEST) measurement model integrating four dimensions of power, energy, form and time sequence is constructed; then, an adaptive graph fusion clustering (MAGFC) unified optimization model is established, each power station is regarded as an independent view, the similarity subgraph, fusion weight, global consensus graph and spectral embedding matrix of each view are automatically learned through optimization, and the optimal clustering number is automatically determined based on graph theory; finally, an alternating direction optimization algorithm is used to solve the model, and representative daily curves and their probabilities are extracted according to the spectral embedding result to form a multi-dimensional typical scene set. The application overcomes the defects of single similarity measurement, poor multi-station cooperation and preset clustering number of traditional methods, and significantly improves the physical fidelity, adaptive ability and engineering practicability of scene extraction.

The application discloses a reasoning type security log analysis method, device, equipment and medium, and relates to the technical field of information security. The method collects multi-source data of a power monitoring system and extracts features; a target knowledge graph is constructed with the ATT&CK framework as the core, an entity state and a correlation relationship are described by using a time sequence quadruple and are converted into a graph network; a reasoning model based on a generative adversarial network is constructed, deep features of the graph network and multi-source data features are spliced and input into a generator, and alternating training is carried out in combination with graph constraint; detected data is input into the converged model to determine abnormalities, attack behaviors are traced back based on a graph structure, and a link is restored. The application fuses physical time sequence logic and deep graph calculation, effectively breaks the model illusion, reduces the false alarm rate, and realizes minute-level accurate tracing.

The application relates to the fields of computer-aided design, knowledge graph and modeling technology, in particular to an engineering drawing change intelligent synchronization method and device, which comprises the following steps: constructing a dynamic semantic topology graph, and mapping components in a three-dimensional model into graph nodes; receiving and analyzing a change instruction for the three-dimensional model; when the change type is a topology reconstruction type change, executing graph reorganization logic based on the dynamic semantic topology graph, identifying affected associated nodes, and automatically deducing a new topology state of the affected nodes according to a preset engineering rule library to generate an updated dynamic semantic topology graph; based on the updated dynamic semantic topology graph, combining a multi-view projection constraint network, calculating target states of two-dimensional graph elements under each view, and rendering and generating a synchronized two-dimensional engineering drawing. The application solves the problem of synchronization failure caused by ID mapping fracture in the prior art, and eliminates the projection contradiction and data inconsistency problems between multi-views caused by independent updating.

The application discloses a patent technology prediction system and method fusing a time sequence knowledge graph and contrast learning, relates to the patent technology prediction field, and is proposed in view of the problem of inaccurate technology prediction in the prior art. Patent elements are extracted and standardized; quadruples are constructed and heterogeneous relationships are defined; time slicing is performed and a graph snapshot sequence is generated; structural flow and heterogeneous graph structures are encoded respectively; time flow is encoded; patent semantics are represented and a unified space is aligned; patent representation and real associated technology keyword representation are explicitly semantically aligned based on a contrast learning enhancement mechanism; coarse retrieval results are obtained; fine retrieval results are obtained and rearranged; coarse ranking scores and fine correction scores are fused and output; multi-level ranking targets are jointly designed; a rearranger and a Gold Injection strategy are trained; and an overall loss function and end-to-end training are performed. The application has the advantages of being capable of depicting the dynamic evolution process of patent technology association changing over time, improving the overall quality of prediction results, and the like.

The application provides a knowledge graph multi-hop question answering method fusing a large model and a graph convolution network, and belongs to the technical field of device troubleshooting. The method comprises the following steps: performing semantic analysis on a question to generate a question hierarchical semantic embedding vector, a feature embedding vector of a core fault entity, and an embedding vector of a potential fault relationship; screening out candidate maintenance step entities with high scores to form an initial candidate answer set; taking the initial candidate answer set as an initial node to obtain graph structure features; fusing the question hierarchical semantic embedding vector, the semantic embedding vector of the candidate maintenance step entity, and the graph structure features of the candidate maintenance step entity to generate a fusion feature of each candidate maintenance step entity; and inputting the fusion feature into a pre-trained answer prediction model to output a final maintenance step entity answer corresponding to the device fault description question. The accuracy and efficiency of the knowledge graph multi-hop question answering are improved.

The application discloses a knowledge graph enhanced retrieval method based on a value subgraph and structure perception, and comprises the following steps: 1) designing a whole framework of a graph enhanced retrieval method based on a value subgraph and structure perception; 2) retrieving a value subgraph, first obtaining a globally optimal value subgraph under the constraints of reward, cost trade-off and connectivity by using a PCST algorithm, and then performing propagation expansion based on similarity from a high correlation seed by using an AdaProp algorithm to obtain a final value subgraph; 3) encoding the value subgraph, encoding the value subgraph into a vector representation containing text semantic information and graph structure information at the same time to generate a graph-level representation; and 4) generating an answer of a large model: taking the graph-level representation as a soft prompt to guide the large model to generate a final answer. The knowledge graph enhanced retrieval technology disclosed by the application can significantly improve the accuracy of a large model knowledge question and answer.

This invention discloses a method and system for constructing an ophthalmological knowledge graph. The invention relates to the technical field of knowledge graphs, and involves labeling a patient's multimodal data in the ophthalmological dimension, combining this data with a corresponding standard ontology for deep learning to output corresponding ontology association paths. Multiple key data points are determined by parsing these ontology association paths. Furthermore, an adaptive optimization mechanism is used to construct an ophthalmological twin, improving its accuracy. The ophthalmological twin is loaded with corresponding ophthalmological constraint relationships, and a corresponding ophthalmological relationship network is constructed using a graph neural network. Ophthalmological anomaly data is simultaneously acquired, and dynamic deduction is performed using the ophthalmological relationship network to construct an N-gram knowledge graph containing temporal relationships. A semantic search mechanism is then used to determine the corresponding target data. Based on the parsing of this target data, multiple factors to be pre-analyzed in different dimensions are determined to determine the updated data combination, thus realizing the dynamic updating of the N-gram knowledge graph.

The application provides a multi-modal knowledge graph representation learning system and product based on subgraph learning, and relates to the technical field of representation learning, and comprises a multi-modal subgraph construction subsystem, which extracts the multi-modal structural information of the head entity in a target triple, and obtains a multi-modal subgraph, wherein the multi-modal subgraph comprises a visual scene graph and a self-centered graph of the head entity; a neighborhood feature aggregation subsystem, which fuses multiple channel feature components of entity features to obtain multi-modal embedding representation of the entity; and a link prediction subsystem, which predicts missing entities and relationships in the target triple based on the multi-modal embedding representation of the entity. The application uses a novel picture structure information extraction mechanism, introduces structured information in the picture by using an efficient graph alignment mechanism, effectively aggregates the neighborhood topological information of the target entity, lays a foundation for efficient fusion of subsequent multi-modal information, and can efficiently learn multi-modal knowledge graph embedding representation.

The application discloses a kind of echarts-based relationship graph drawing method and device, the method includes: defining the node data structure of node in relationship graph, define canvas structure and initialize relationship graph drawing canvas according to the canvas structure, determine graph center coordinates;According to the node data structure, construct node, determine the connection relationship between the node;Based on echarts, according to the graph center coordinates, the connection relationship between the node is drawn and generated visual relationship graph;The visual relationship graph includes upper visual relationship graph and center visual relationship graph.Through the application, can make the data of only last two layers in graph, that is, there are at most two center nodes in graph, the name of each center node in complete association path is displayed in the form of text above graph, both reduce the data amount of graph rendering, and do not affect the depth relationship query of data and the presentation of complete relationship path.

This invention discloses a method and apparatus for quality assessment and tracing of nodes in aerospace data circulation networks, relating to the fields of artificial intelligence and data processing technology. First, a standardized aerospace data tracing knowledge graph is constructed, unifying the modeling of three types of entities—data, processing, and subjects—and three types of relationships—production, execution, and input. Second, multimodal feature encoding and heterogeneous graph reconstruction are carried out, fusing quality attributes and topological structure to form a computational graph. Then, a quality regression model is trained based on the heterogeneous graph to achieve continuous numerical prediction of node quality. Finally, combining upstream path constraints and multi-dimensional attribution mechanisms, the root causes of quality problems are located and an interpretable diagnostic report is output. This invention transforms quality analysis from discrete discrimination to continuous quantitative assessment, accurately adapting to heterogeneous network semantics, achieving integrated assessment and tracing, significantly improving the accuracy, efficiency, and interpretability of aerospace data quality analysis, and is applicable to aerospace data end-to-end quality monitoring and reliable circulation scenarios.