86 results about "Schema graph" patented technology

The present disclosure relates to systems, non-transitory computer-readable media, and methods for linking a database schema to a natural language query. In particular, in some embodiments, the disclosed systems determine, from tables in a database schema, a subset of tables relevant to a natural language query by comparing embeddings for the tables in the database schema and embeddings for the natural language query. Additionally, in some implementations, the disclosed systems select, from a schema graph comprising nodes that represent the tables in the database schema, an additional table along a path between a pair of nodes representing a pair of tables from the subset of tables. Moreover, in some embodiments, the disclosed systems determine a set of relevant tables by appending the additional table to the subset of tables. Furthermore, in some implementations, the disclosed systems generate, from the set of relevant tables, a response for the natural language query.

The invention relates to a Text-to-SQL (Structured Query Language) complex query statement generation method and system based on a pre-trained large model. The method comprises the following steps: acquiring an input sequence comprising a user question and dictionary interpretation of table names and column names in a database; respectively inputting the user question and the dictionary explanation into a Sension-BERT model, obtaining Top-K explanations most related to the user question, and obtaining an enhanced input sequence according to the Top-K explanations most related to the user question; encoding the enhanced input sequence to generate an enhanced semantic vector, and constructing a hierarchical database mode pattern according to the enhanced semantic vector; inputting the hierarchical database mode graph into a hypergraph neural network model to update node features, and obtaining optimized node features; and obtaining a structured query statement according to the optimized node features. According to the method, complex association and multi-level query logic among multiple tables in a complex query scene can be better processed, and the performance in the complex query scene is improved.

The invention discloses a motor fault diagnosis method and system based on color image fusion symmetric point mode, and the method comprises the steps: converting a vibration signal and an electromagnetic signal of a motor into symmetric point mode images, and generating a color signal image fusing feature information; respectively abstracting the color signal images fused with the feature information into nodes and edges in a high-dimensional semantic space so as to construct graph structure data; and performing diagnosis classification on the graph structure data of the vibration signals and the electromagnetic signals by using respective capsule graph network models, and fusing diagnosis classification results of the vibration signals and the electromagnetic signals through a voting mechanism to obtain a final diagnosis classification result. According to the method, multi-channel time domain signals are converted into image expressions with dense information and consistent geometry, and unified feature modeling is carried out on the images based on a depth map structure network with topology perception capability, so that motor fault diagnosis with high diagnosis precision and strong robustness is realized.

The invention belongs to the technical field of compressor maintenance, and discloses a compressor predictive maintenance system and method based on multi-source data fusion, and the method comprises the steps: carrying out the heterogeneous data time-frequency feature analysis of compressor signal data, and generating a multi-dimensional feature spectrum; carrying out load-dependent fault signal separation to generate load decoupling fault feature data; performing environment interference factor elimination analysis to generate a pure fault feature matrix; performing cross-domain signal correlation mapping to generate a multi-source data fusion mode map; constructing a fault feature propagation link, and generating fault evolution path data; carrying out degradation trend prediction under a variable load condition, and generating fault development situation data; constructing a dynamic threshold self-adaptive early warning model, and generating fault early warning critical value data; health state comprehensive evaluation is carried out, a compressor state report is generated, and predictive maintenance implementation is executed; according to the invention, accurate recognition and prediction of the compressor fault are realized, and the equipment reliability and maintenance efficiency are improved.

A method of clustering patterns of an integrated circuit includes; providing a pattern image and numeric data, as input data corresponding to a first pattern to a first model, wherein the first model is trained by a plurality of sample images and a plurality of sample values, obtaining a content latent variable using the first model, and grouping a plurality of content latent variables corresponding to a plurality of patterns into a plurality of clusters based on a Euclidean distance, wherein the numeric data represents at least one attribute of the first pattern.

The invention discloses a method for identifying an arcing mode of a vacuum arc based on deep learning. The method comprises the following steps: S1, acquiring an arcing mode image sequence; s2, establishing a deep learning network and obtaining a network training model; s3, inputting an arc image needing to be identified into the training model for identification; the deep learning network selects an improved GoogleNet network in a convolutional neural network, and the improved Googlenet network structure comprises an input layer, an initial convolutional layer, a pooling layer, an inception module, an auxiliary classifier, a global average pooling layer, a Dropout layer, a full connection layer, a softmax layer and an output layer which are connected in sequence; the vacuum arc identification method based on deep learning mainly aims at solving a series of problems that in the prior art, arcing experiments have a large amount of data including arc dynamic images shot by a high-speed camera, and manual identification is long in time consumption, low in efficiency, high in subjectivity and the like.

The invention discloses a large model query generation system and method combining graph structure analysis and execution, and the system comprises a graph mode standardization module which converts a description file into a unified structured representation; the large language model module is used for screening out nodes, edges and sub-graph modes required by query statements according to the natural language problem and the structured representation of the user; the graph mode detection module is used for detecting and modifying the screened nodes, edges and sub-graph modes through scripts; the large language model module generates an initial query statement according to a natural language problem of a user and detected and modified nodes, edges and subgraph modes; the query statement detection module is used for detecting the initial query statement through a script and generating an error report; and the large language model module generates a modified query statement. According to the large model query generation system and method combining graph structure analysis and execution, the accuracy of graph mode selection is remarkably improved through the thinking ability of the large language model and the script error detection ability, and therefore the quality of graph database query statement generation is improved.

The invention relates to the technical field of medical treatment, and discloses a knee joint knee protection treatment personalized planning system based on multi-modal data fusion. A data acquisition and preprocessing module of the system acquires a multi-modal data stream of the knee joint of a patient in real time and generates a standardized data packet; the treatment event identification module is used for automatically identifying and verifying key treatment events from the standardized data packet and generating a treatment event sequence; the historical treatment path query module queries a distributed database and extracts treatment modes of similar patients; the graph matching and path generation module is used for calculating the similarity between the current treatment event sequence and a historical treatment mode, generating a fusion treatment path when the similarity exceeds a threshold value and storing the fusion treatment path to the block chain network; the abnormity monitoring and analysis module monitors abnormal event points in the fusion treatment path in real time, triggers a path decomposition mechanism and performs multi-dimensional performance analysis; and the personalized scheme generation module is used for dynamically adjusting treatment parameters and generating a personalized treatment planning scheme.

The invention discloses a network attack behavior detection method and device, a storage medium and computer equipment, and the method comprises the steps: obtaining a plurality of pieces of network data, and constructing a to-be-analyzed graph structure based on the plurality of pieces of network data; based on a preset attack mode pattern, judging whether a first target sub-graph matched with the preset attack mode pattern exists in the to-be-analyzed graph structure, and when the first target sub-graph exists, determining that a network attack behavior indicated by the preset attack mode pattern exists in the to-be-analyzed graph structure; and generating a network attack behavior detection result corresponding to the plurality of pieces of network data according to the first target sub-graph and the corresponding network attack behavior. According to the invention, the attack behavior in the network can be detected more comprehensively and more accurately, and the possibility of false report and missing report is reduced; in addition, potential network attack behaviors can be quickly positioned, the detection efficiency is improved, and timely response to network security threats is facilitated.

The invention relates to the technical field of query characterization, and discloses a database query feature vector generation method, which comprises the following steps of: obtaining a connection bitmap and a database mode graph of a database queried by a target query statement; determining a database mode sub-graph related to the target query statement according to the database mode graph; encoding the database mode sub-graph through a preset gating graph neural network to generate a mode item vector; and determining a query feature vector corresponding to the target query statement according to the connection bitmap and the mode item vector. According to the method, by obtaining the database mode graph and the connection bitmap, the structure association information between the table and the column in the database and the predicate condition features of the query can be synthesized, the semantic association represented by the features is enhanced based on the coding capability of the graph neural network, and the expression capability of the query feature vector for query intention and data association is effectively improved.

The invention relates to the technical field of power grid side energy storage, in particular to a dynamic evaluation and optimization method for power grid side energy storage system site selection, which comprises the following steps of: based on a power grid side demand, executing power grid load data analysis, calculating a difference value between a peak load and a minimum load, comparing with historical load data, identifying a fluctuation mode of the power grid demand, and calculating the power grid side energy storage system site selection. And obtaining a load fluctuation mode pattern. According to the invention, by monitoring and analyzing the change of the power grid load in real time, accurately adjusting the deployment of the energy storage equipment and using the long-short-term memory network to optimize the prediction accuracy, the decision error is significantly reduced, the selection of the position and scale of the energy storage system is more scientific, and by combining the cost, the reliability and the environmental influence, the prediction accuracy is improved. The multi-objective optimization enhances the comprehensiveness of the decision, the integration of the geographic information system optimizes the processing of the spatial data, the adaptability and implementation efficiency of the site selection strategy are improved, and the decision is ensured to be implemented under the definite constraint condition, so that the higher operation efficiency and utilization rate of the energy storage system are realized.

The invention discloses a method and a system for detecting a complex multi-step attack based on an interpretable graph neural network, and a storage medium. The method comprises the following steps: generating a strong negative sample by using priori attack knowledge; inputting the overall graph into a detection model to obtain an attack detection result; and inputting the attack event and the detected attack behavior into an interpreter to obtain the interpretation of the attack behavior. Constructing a traceability graph and an attack mode graph from a system log, aligning the generated traceability graph and attack mode graph, and enhancing an overall graph; pre-training an encoder of a multi-step attack detector on the continuous time dynamic heterogeneous graph by using comparative learning; finely adjusting the model on a small number of real attack samples based on a pre-training model; the relevance between an abnormal event and a preorder event is calculated by using a sniffer, and then an interpretable sub-graph is searched and output by using a Monte Carlo tree under the guidance of the sniffer by a digger. The problems that an existing method is difficult to solve the problem of data imbalance, low in interpretability and the like are solved.

The invention provides an intelligent fault positioning system and method based on a DC system, and relates to the technical field of power fault positioning, and the system comprises a digital twin module, a sensor network module, a real-time data synchronization module, a graph neural network model module, an online reasoning module, an intelligent decision module and a decision self-healing module. Through deep fusion of the digital twinning technology and the graph neural network, complex electrical, thermal, mechanical and environmental parameters in the direct current system can be comprehensively analyzed. The digital twin module provides a high-precision virtual model of the system through multi-physics field integration, so that a graph neural network model can accurately capture tiny anomalies and complex fault modes in the system under the support of rich and high-quality data. Through combination of the graph convolutional network and the graph attention network, the graph neural network model can effectively identify relevance between key nodes and edges, weights of neighbor nodes are dynamically adjusted, and the fault detection and positioning precision is further improved.

The invention provides a cooperative identification method and system for a power line communication signal and an arc fault, and the method comprises the steps: firstly, constructing and training a double-flow communication fault sensor comprising a sensing early-warning sub-module and an active sensing sub-module, then carrying out the communication fault sensing based on the double-flow communication fault sensor, generating a communication fault sensing data set, and transmitting the communication fault sensing data set to a server; then, a graph neural network is adopted to carry out fault mode recognition on the communication fault sensing data set, a communication fault mode graph is generated, then, communication fault positioning is carried out based on the communication fault mode graph, a communication fault positioning data set is generated, and finally, actual feedback data of the communication fault positioning data set is obtained. And the system is updated according to the actual feedback data, so that the fault diagnosis rate and accuracy are improved.

The invention relates to an edge calculation-based equipment fault diagnosis method and system for a battery swap station. The method comprises the following steps of: acquiring data and constructing a regional equipment model; deploying the regional equipment model at an edge computing node, and collecting operation time sequence data in real time; obtaining expected state data through the lightweight simulation model, and comparing the expected state data to generate a residual sequence; obtaining a causal violation event set through a lightweight causal graph model; generating a fault diagnosis result through the fault mode map; in conclusion, the regional equipment model comprising the lightweight causal graph model, the simulation model and the fault mode graph is constructed, real-time data acquisition and closed-loop correction are realized in combination with the edge computing nodes, and the dynamic threshold optimization and fault mode graph fusion mechanism is utilized. The problems that a traditional method is large in response delay, high in false alarm rate and insufficient in diagnosis precision are effectively solved, and the effects of improving fault diagnosis real-time performance, reducing cloud data transmission delay, improving diagnosis precision and reducing the false alarm rate are achieved.

The invention provides a multi-modal knowledge graph construction and question answering method and system based on query driving, and relates to the field of knowledge graphs, and the method comprises the steps: constructing a mode graph of user query; constructing a multi-modal knowledge base; constructing a clue map through a mode constraint retrieval method and a query focusing clue refining method in combination with the mode map; and through a cross-modal reasoning iteration mode, supplementing missing information of the cue graph to obtain a retrieval answer result queried by the user. According to the method, a query-driven multi-modal framework is constructed, the local knowledge graph of user query can be dynamically constructed, and rapid and accurate retrieval of user query problems is realized.

The invention discloses a coarse-grained reconfigurable architecture cyclic mapping method and device based on a graph neural network, and relates to the technical field of computer reconfigurable computing. The method comprises the following steps: based on a priority prediction model, performing list scheduling according to a data flow diagram and hardware information to obtain a list scheduling table; performing modular scheduling according to the data flow diagram and the hardware information; pre-scheduling the data flow diagram on the basis of a list scheduling table according to the modular data flow diagram and the time extension coarse-grained array to obtain a time step attribute data flow diagram; obtaining a mapping result by using a mapping algorithm based on pattern diagram matching according to the time step attribute data flow diagram on the basis of the time extension coarse-grained array; based on the hardware information, processing unit isomorphism verification is carried out according to the mapping result; and compiling the verified mapping result to obtain an executable configuration information file. The invention provides an efficient and accurate coarse-grained reconfigurable architecture cyclic mapping method based on a graph neural network.

The invention relates to the technical field of data analysis, and discloses an electronic terminal MTBF analysis method and device based on an AI model, equipment and a medium, and the method comprises the steps: carrying out an MTBF test through a pre-trained AI test model according to a pre-constructed virtual parameter set, and obtaining an MTBF test report, identifying fault time distribution in the virtual parameter set according to the MTBF test report, identifying a reliable parameter set and a fault-prone parameter set in the virtual parameter set based on a preset threshold value, and constructing a fault mode map according to the virtual parameter set and the MTBF test report, and identifying a sensitive coefficient of each parameter type according to a virtual parameter set and the fault mode atlas based on parameter disturbance, generating a reliable parameter range according to the reliable parameter set, and generating a parameter adjustment suggestion list for each virtual parameter in the virtual parameter set in combination with the sensitive coefficient and the reliable parameter range. And the accuracy of the MTBF test is improved.

Aspects of the disclosed technology provide solutions for extracting subgraph patterns in graph-structured data and encoding them as embeddings using a graph neural network (GNN). In some aspects, a process of the disclosed technology can include steps for receiving an input graph comprising a plurality of nodes and edges, the input graph representing relationships among a plurality of entities, parameterizing a graph neural network model based on a set of pattern graphs, and identifying, for at least a portion of the nodes in the input graph, rooted homomorphisms between the pattern graphs and local subgraphs rooted at the respective nodes. Systems and machine-readable media are also provided.

The present invention belongs to the technical field of energy management, and discloses a building energy management system based on intelligent gateway edge computing; the method comprises: collecting energy usage data and environmental data inside a building; transmitting the collected energy usage data and environmental data to an edge computing node through an intelligent gateway; in the edge computing node, preprocessing the collected energy usage data and environmental data to obtain preprocessed data; based on the preprocessed data, constructing an energy usage pattern diagram, optimizing the structure of the energy usage pattern diagram, and identifying key nodes and key associations; according to the identified key nodes and key associations, formulating an energy optimization strategy using an optimization algorithm, sending the energy optimization strategy to the intelligent gateway, and controlling the energy usage equipment inside the building through the intelligent gateway, so as to realize efficient utilization of energy, improve the intelligence and automation level of building energy management, and reduce management difficulty and labor cost.

The invention discloses a transaction behavior auditing method and device, equipment and a medium, and relates to the technical field of financial information, and the method comprises the steps: determining a target violation mode graph based on a mode query instruction, determining a time window and a matching search tree, and determining target flow data based on the time window; constructing a data graph by taking the transaction participants as vertexes and the transaction behaviors as edges, and determining to-be-matched edges; matching the to-be-matched edge with a target violation mode graph, and after a target matched edge exists, pruning and updating the matched search tree by using a preset pruning strategy to determine whether the current to-be-matched edge meets a time constraint condition of the target violation mode graph; and judging whether a violation mode is detected based on a determination result, and skipping to the step of determining the edge to be matched until a plurality of edges of the data graph are matched. Therefore, the problem of low auditing efficiency caused by difficulty in identifying violation operation with time constraint due to the fact that a complex sequential relationship between transactions cannot be captured can be solved.

The present invention relates to the field of privacy protection technology, and discloses a method for publishing the number of range subgraphs based on differential privacy, including inputting a point attribute graph and privacy parameters; sorting the nodes in ascending order according to the value of the attribute; initializing an intermediate vector; enumerating all copies of the pattern graph in the point attribute graph; constructing a differential privacy range tree with all Euclidean space points as input; obtaining a new discretization boundary value according to the discretization range of each query; and recursively traversing the differential privacy range tree generated in step four according to the attribute range to obtain the number of copies of the pattern graph in the point attribute graph within the attribute range. The present invention can realize an efficient ε-DP method with the same additive error for the range subgraph counting problem with edge attributes through simple transformation. The method of the present invention is concise and easy to deploy, and its time complexity is consistent with that of the non-private range subgraph counting method.

Systems and methods are disclosed for identifying target graphs that have nodes or neighborhoods of nodes (sub-graphs) that correspond with an input query graph. A visual analytics system supports human-in-the-loop, example-based subgraph pattern search utilizing a database of target graphs. Users can interactively select a pattern of nodes of interest. Graph neural networks encode topological and node attributes in a graph as fixed length latent vector representations such that subgraph matching can be performed in the latent space. Once matching target graphs are identified as corresponding to the query graph, one-to-one node correspondence between the query graph and the matching target graphs.

The present disclosure relates to systems, non-transitory computer-readable media, and methods for linking a database schema to a natural language query. In particular, in some embodiments, the disclosed systems determine, from tables in a database schema, a subset of tables relevant to a natural language query by comparing embeddings for the tables in the database schema and embeddings for the natural language query. Additionally, in some implementations, the disclosed systems select, from a schema graph comprising nodes that represent the tables in the database schema, an additional table along a path between a pair of nodes representing a pair of tables from the subset of tables. Moreover, in some embodiments, the disclosed systems determine a set of relevant tables by appending the additional table to the subset of tables. Furthermore, in some implementations, the disclosed systems generate, from the set of relevant tables, a response for the natural language query.

The invention provides a zero sample standing field detection method based on cognitive mode graph enhancement. The method comprises the following steps: constructing a logic cognitive mode; converting an input text into a logic expression by using a pre-trained large language model LLM; clustering logic predicates in the logic expression, and organizing the logic predicates into different clusters; summarizing the texts in each cluster to generate summarized phrases of the cluster; constructing a weighted multi-relation graph as a cognitive mode according to the logic relation between the logic predicates and the summary phrases of the clusters; the method comprises the following steps of: predicting a cognition enhanced vertical field: constructing a cognition-first-order logic rule graph CFGraph; carrying out modeling on the constructed first-order logic rule graph CFGraph by utilizing a relational graph convolutional network so as to carry out vertical field prediction; and selecting the category with the highest probability as a prediction result. The method aims at reducing dependence on annotation data, improving the generalization ability of the model and enhancing the field adaptability of the LLM.

This application provides a method for processing tabular question-and-answer data and training a model, an electronic device, and a storage medium. The method for processing tabular question-and-answer data includes: obtaining a first representation vector corresponding to a natural language query statement; obtaining a pattern item vector corresponding to the first representation vector based on a question-pattern graph, wherein the question-pattern graph is used to represent the association between word segmentation in the natural language query statement and database pattern items, and the question-pattern graph is pre-generated at least based on a semantic association question-pattern subgraph, which is generated based on the semantic association between natural language query statement samples and corresponding database pattern item data samples obtained after masking; obtaining a second representation vector based on the first representation vector and the pattern item vector, and generating a corresponding database query statement based on the second representation vector.

The present disclosure relates to a static graph mode graph generation system supporting dynamic resolution. The system comprises: an original logic graph construction component, for a first specific graph generation task, tracking original computing logic by executing a construction function of a graph to construct an original logic graph; a logic graph optimization component, forming a first logic graph for the constructed original logic graph; a node insertion component, inserting a communication node and a runtime control node required by distributed computing for the first logic graph to generate a first runtime logic graph for multi-device collaboration; a runtime plan generation component, generating a first runtime execution plan based on a data dependency relationship of the first runtime logic graph; and a runtime initialization component, initializing a first specific graph runtime instance according to the first runtime execution plan to complete dynamic allocation of computing resources and configuration of an execution environment.

This invention discloses a method for identifying damage patterns in carbon fiber reinforced composite materials using an interpretable convolutional neural network. The method comprises the following steps: establishing a database; preprocessing data; constructing a convolutional neural network model; interpreting the convolutional neural network model; and generalizing the convolutional neural network model. After model training, the model is generalized and applied to identify multiple damage pattern images and composite impact damage patterns in actual engineering projects to verify the robustness of the model. The technical solution provided by this invention can accurately and efficiently identify damage patterns in carbon fiber reinforced composite materials, achieving precise quantification and efficient decoupling of composite material damage patterns under service conditions.

The present application provides a query-driven multimodal knowledge graph construction and question-answering method and system, which relates to the field of knowledge graphs. The method includes: constructing a pattern graph of user queries; constructing a multimodal knowledge base; combining the pattern graph with a pattern constraint retrieval method and a query focus clue refinement method to construct a clue graph; supplementing the missing information of the clue graph through cross-modal reasoning iteration to obtain the retrieval answer result of the user query. The present application constructs a query-driven multimodal framework that can dynamically construct a local knowledge graph of user queries, enabling fast and accurate retrieval of user query questions.

The invention belongs to the technical field of medical information and artificial intelligence, and discloses a health trajectory prediction method and system based on double-branch collaborative learning. According to the sequence-graph double-branch collaborative architecture provided by the invention, the limitation of a single modeling view angle is fundamentally changed. The sequence branch focuses on mining a deep time sequence evolution mode in a multi-mode treatment sequence, and the graph branch depicts dynamic topological association among treatment events in a fine manner. The two are fused through a self-adaptive mechanism, high-quality dynamic representation capable of comprehensively reflecting time sequence evolution and structure correlation of a patient is generated, and a solid foundation is laid for accurate prediction.