28 results about "Graph similarity" patented technology

The invention provides a three-dimensional part retrieval method based on graph similarity search, relates to the field of computer graphics, and solves the technical problems that in the prior art, geometric and design semantic information of CAD cannot be fully utilized, and a topological relation is difficult to capture and display, so that the retrieval efficiency is low. The method comprises the following steps: acquiring part three-dimensional data of a computer-aided design (CAD) model, and constructing a training data set; constructing an edge-surface connection diagram based on the three-dimensional data of the part; calculating a graph editing distance (GED) matrix of all edge surface connection graphs in the training data set as a supervision signal; based on the GED matrix, training a sorting model, mapping an edge-surface connection graph to a hidden space, and constructing a part vector database according to an output graph-level embedding vector; the sorting model is constructed based on a graph attention network; and inputting a to-be-queried CAD part into the trained sorting model to obtain a feature vector, carrying out nearest neighbor search in the part vector database, and returning a similar part result.

The application provides a three-dimensional part retrieval method based on graph similarity search, relates to the field of computer graphics, and solves the technical problem that the prior art cannot fully utilize the geometry and design semantic information of CAD, and it is difficult to capture and display the topological relationship, resulting in low retrieval efficiency. The method comprises the following steps: obtaining part three-dimensional data of a computer-aided design (CAD) model, and constructing a training data set; constructing an edge-face connection graph based on the part three-dimensional data; calculating a graph edit distance (GED) matrix of all edge-face connection graphs in the training data set as a supervision signal; training a ranking model based on the GED matrix, mapping the edge-face connection graph to a hidden space, and constructing a part vector database according to the output graph-level embedding vector; the ranking model is constructed based on a graph attention network; inputting a CAD part to be queried into the trained ranking model to obtain a feature vector, performing nearest neighbor search in the part vector database, and returning a similar part result.

The invention provides a regulation query method and device based on a coal mine regulation knowledge graph, and relates to the technical field of coal mine safety. The method comprises the steps of obtaining a to-be-queried coal mine case text, a coal mine case set and a coal mine regulation knowledge graph; performing semantic similarity calculation on the case text and each historical case, and taking the obtained semantic similarity as a semantic similarity component of a corresponding regulation; calculating the graph similarity between the case text and each node of the knowledge graph to obtain the graph similarity component of each regulation; and fusing the semantics and the graph similarity component to select a target regulation and generate a reply text. According to the method, semantic matching and graph structure association are fused, so that target terms which are related to case text semantics and are closely structurally associated can be accurately screened from an overall regulation system, and end-to-end accurate query from case description to regulation matching is realized; and the accuracy and integrity of regulation basis acquisition in coal mine safety supervision are obviously improved.

The present disclosure provides a knowledge graph-based industrial knowledge mining modeling method and system, relating to data mining technology, the method comprising: obtaining basic information of a pre-constructed knowledge graph; determining data collection type and data collection amount; collecting relevant data on multiple same-type equipment products of a target device to construct multiple target knowledge graph sets; based on device usage time nodes, traversing and comparing the multiple target knowledge graph sets to obtain a graph similarity feature set; constructing a target device twin model and embedding graph similarity features in the graph similarity feature set into the target device twin model according to device usage time nodes; and performing fault early warning on the target device usage process according to the target device twin model. The present disclosure can solve the technical problem of low accuracy and timeliness of existing industrial equipment fault early warning methods, and can improve the accuracy and efficiency of industrial equipment fault early warning, thereby timely discovering and handling fault hazards.

The invention relates to a similarity calculation method of a affair graph and a related product. The method comprises the following steps: acquiring a plurality of affair maps; based on different types of information similarity models, respectively analyzing the similarity of the target information between the affair maps to obtain different types of information similarity features; on the basis of domain weights, the information similar features are fused, associated features are obtained, and the domain weights are determined according to domain knowledge; fusing the weight of each matching edge under the maximum weight matching of the bipartite graph to obtain a graph similarity; the bipartite graph is determined according to the associated features. According to the scheme provided by the invention, the domain weight is determined through the domain knowledge, and the fusion process of the information similar features is guided to improve the influence of key elements in the information similar features on the associated features, so that the related parameters of the bipartite graph more accurately reflect the similarity degree between the two affair maps, and accurate calculation of the map similarity is facilitated.

The invention relates to generating violation exemption terms for static lint checks. In one example, a method includes converting source code for register transfer level design into a directed graph; acquiring a first violation generated by analyzing the source code; identifying a violation statement sub-graph associated with the first violation in the directed graph; extracting a reduced sub-graph representing the first violation from the directed graph, wherein the violation statement sub-graph comprises a starting point of the reduced sub-graph; converting the reduced sub-graph into a first vector; calculating a graph similarity between the first vector and a second vector representing a second violation of the source code for which an existing exempt item has been generated; determining, by a processing device, that the graph similarity satisfies a threshold similarity; and generating an exemption item for the first violation in response to the determination.

The application discloses a residential building function relationship graph recommendation method based on graph similarity and graph edit distance, and comprises the following steps: 1) establishing a residential function relationship graph database; 2) obtaining a function relationship graph input by a user; 4) obtaining a function relationship graph to be matched; 5) searching for residential function relationship graphs with the same number of vertices as the function relationship graph to be matched, and respectively calculating the similarity between the function relationship graph to be matched and the residential function relationship graphs; 6) based on the similarity, arranging the searched residential function relationship graphs in descending order, and then calculating the graph edit distance between the function relationship graph to be matched and the residential function relationship graphs; 7) based on the graph edit distance, arranging the searched residential function relationship graphs in ascending order, and selecting K residential function relationship graphs as recommended residential function relationship graphs. The application realizes the recommendation of the residential building function relationship graph, the recommended scheme is accurate and diverse, can be used in various building auxiliary design software, and has high portability.

The disclosure provides a regulation query method and device based on a coal mine regulation knowledge graph, and relates to the technical field of coal mine safety. The method comprises the following steps: obtaining a coal mine case text to be queried, a coal mine case set and a coal mine regulation knowledge graph; calculating the semantic similarity of the case text and each historical case, and taking the obtained semantic similarity as the semantic similarity component of the corresponding regulation; calculating the graph similarity of the case text and each node of the knowledge graph to obtain the graph similarity component of each regulation; and fusing the semantic and graph similarity components to select a target regulation and generate a reply text. Through the fusion of semantic matching and graph structure association, the disclosure can accurately screen the target clauses that are closely related to the semantic of the case text and the structure from the overall regulation system, realizes end-to-end accurate query from case description to regulation matching, and significantly improves the accuracy and completeness of the regulation basis obtained in the coal mine safety supervision.

In one example, a method includes converting source code for a register transfer level design into a directed graph, acquiring a first violation generated by analyzing the source code, identifying a violation statement subgraph associated with the first violation in the directed graph, extracting a reduced subgraph representing the first violation from the directed graph, wherein the violation statement subgraph comprises a starting point for the reduced subgraph, converting the reduced subgraph to a first vector, calculating a graph similarity between the first vector and a second vector representing a second violation for the source code for which an existing waiver has been generated, determining, by a processing device, that the graph similarity satisfies a threshold similarity, and generating, in response to the determining, a waiver for the first violation.

The application belongs to the field of protein map prediction, and discloses a single-sequence protein contact map prediction method based on a graph neural network, which comprises the following steps: collecting protein sequences and corresponding protein three-dimensional structure data, and constructing a residue contact matrix; performing one-hot coding processing on each protein sequence to obtain one-hot one-dimensional features; obtaining one-dimensional protein coding features and two-dimensional attention matrix features of each protein sequence through a pre-training model; constructing a graph convolution coding model and a convolution decoding model; inputting the one-hot one-dimensional features and the one-dimensional protein coding features obtained through pre-training as input features into the graph convolution coding model to obtain protein structure coding features; inputting the protein structure coding features and the two-dimensional attention matrix obtained through pre-training into the convolution coding model to obtain a residue pair contact prediction matrix in the protein sequence. The protein contact map generated by the method has high similarity with a real contact map and good prediction accuracy.

The application discloses a digital person virtual person setting generation method, and relates to the technical field of data processing. The method comprises the following steps: in response to an input task theme, a target region identifier and an industry classification identifier, generating a person setting genotype structure according to the target region identifier and the industry classification identifier; constructing a generation environment control file based on the person setting genotype structure; loading the generation environment control file into a pre-trained generation model to generate differentiated content according to the task theme; analyzing the cognitive homogenization risk of the differentiated content through a graph isomorphism detection algorithm; and if the graph similarity of the cognitive homogenization risk exceeds a preset threshold, updating the person setting genotype structure. The application reduces the homogenization degree of the content output by a large model.

The invention relates to an APT (Advanced Persistent Threat) attack analysis technology based on LLM (Logical Language Model) and RAG (Random Access Gateway), which is used for automatically analyzing an APT attack detection result and generating an APT survey report in a natural language form. The automatic APT attack analysis system provided by the invention comprises the following three key subsystems: 1, a CTI information extraction system based on LLM intelligent agent structure alignment: using an atlas extractor and an atlas inspector intelligent agent to automatically construct a CTI atlas structurally aligned with a traceability graph from CTI according to a predefined mode; and 2, an RAG system based on a GNN retrieval model: using a graph similarity matching model to retrieve a CTI graph corresponding to the traceability graph. And 3, an APT attack analysis system capable of autonomously and dynamically planning a reasoning process is used, so that traceable and automatic APT attack analysis of the reasoning process is realized. The method can be used for network attack analysis including but not limited to APT attacks, and the efficiency and accuracy of network attack analysis can be effectively improved.

The application provides an event integration method, device and equipment and a computer readable storage medium, which are applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation and vehicle-mounted devices. The method comprises the following steps: obtaining an event to be integrated, obtaining at least one topic to be integrated, each topic to be integrated comprising at least one topic event; based on selection logic, selecting one or more of semantic similarity, string graph similarity and question and answer similarity to obtain a target similarity between the event to be integrated and each topic to be integrated, the string graph similarity being the similarity of a graph feature corresponding to a key string, and the question and answer similarity being the similarity of a question and answer feature; based on the target similarity, determining a target topic to which the event to be integrated belongs from the at least one topic to be integrated; and integrating the event to be integrated into the target topic to obtain an event context comprising the event to be integrated and the at least one topic event. Through the application, the accuracy of event integration can be improved.

The invention discloses a drug screening-oriented small molecule graph similarity calculation and retrieval method and device. Comprising the following steps: analyzing a chemical structure of a small molecule compound, converting into a standard small molecule graph, and constructing a graph pair data set; extracting structure and semantic information in the graph by utilizing a graph isomorphic network, and generating node and edge embedding; based on node and edge embedding, semantic similarity and structure consistency information are fused, and an alignment result is calculated; performing embedding exchange fusion and secondary coding, and enhancing embedding re-interaction to obtain alignment cost; carrying out model training by utilizing a mean square error of a true value and a predicted value; on the basis of an anchor point atlas selected iteratively and node embedding pooling, generating graph embedding containing global features; constructing a weighted HNSW neighbor graph index according to graph embedding and positive and negative samples; and carrying out rough arrangement and fine arrangement collaborative retrieval by utilizing a weighted HNSW neighbor graph index and a graph similarity calculation model to obtain the most similar K small molecules. The efficiency and precision of small molecule drug screening are improved.

A method for training a model for determining graph similarity is disclosed. The method comprises receiving a first graph and a second graph as training inputs, the first graph and the second graph each including nodes connected by edges. The method further comprises applying a model to the first graph and the second graph to determine (i) pairs of aligned nodes between the first graph and the second graph and (ii) a first training loss. The method further comprises generating a first augmented graph by modifying the first graph depending on the pairs of aligned nodes. The method further comprises applying the model to the first graph and the first augmented graph to determine a second training loss. The method further comprises refining the model based on the first training loss and the second training loss.

The invention belongs to the technical field of network security detection, and provides an artificial intelligence-driven network intrusion detection method, system and device. The method comprises the steps of multi-source metadata extraction and preprocessing, causal feature discovery and selection, meta learning detection model reasoning, known attack screening, causal comparative analysis and branch judgment, adaptive threshold adjustment and interpretable report generation. According to the method, initial parameters of three layers of MLP are optimized by adopting an MAML framework, so that the model can be quickly adapted only by a small amount of fine tuning in a zero-day attack and few-sample scene, and the limitation that traditional machine learning depends on a large number of historical samples is broken through; meanwhile, by constructing a baseline normal causal graph and an abnormal causal graph, adopting GED to quantify the similarity of the two graphs and accurately distinguishing zero-day attack and concept drift based on a preset threshold value, the industrial pain point that zero-day attack detection is difficult is solved, misinformation caused by concept drift is avoided, it is ensured that resources are only used for real attack analysis, and the detection efficiency is improved.

This invention discloses a domain-adaptive data governance standard recommendation system based on a multimodal large model and NLU-HGSC. The invention includes a standard management module, a directory preprocessing module, an NLU-HGSC module, and a data storage module. The standard management module is used for collecting and parsing standards. The directory preprocessing module is used for collecting and preprocessing metadata to be matched. The NLU-HGSC module is used to calculate the similarity between heterogeneous graph structures generated by the standard management module and the directory preprocessing module, and recommend corresponding data standards for the directory data in the data storage module as the basis for data governance. The NLU-HGSC heterogeneous graph similarity calculation technology in this invention requires no training; even if standards are added or modified, the recommendation results will adaptively adjust without requiring model retraining as the standard data changes.

The application discloses a kind of compound structure graph similarity query method based on partition, including obtaining the compound structure graph to be queried, compound structure graph database G, when G is empty, then output similarity result set, when G is not empty, then successively take out compound structure graph g in G, calculate the first edit distance lower bound of g and q and judge whether threshold value is satisfied, if yes, then judge whether G is empty, if no, then set partition end condition, respectively to q and g are partitioned, calculate the second edit distance lower bound of g after partition and q and judge whether threshold value is satisfied, if yes, then judge whether G is empty, if no, then obtain mapping list generated in partition process, according to mapping list, construct state space tree, according to state space tree, calculate the third edit distance lower bound of g and q and judge whether threshold value is satisfied, if yes, then judge whether G is empty, if no, then g is added in similarity result set. Improve the speed and detection precision of similar compound structure graph.

This application provides an artificial intelligence-based robot control method and system, relating to the field of intelligent robots, and solves the technical problems of low execution efficiency and weak safety of existing robots in unfamiliar environments. The method includes: collecting visual data, depth data, 3D point cloud data, and audio data of the robot's environment; processing each modal data to obtain multimodal feature vectors and event labels; constructing dynamic entities representing environmental elements based on the multimodal feature vectors; assigning a multidimensional semantic vector to each dynamic entity to generate a context graph describing the relationships between entities; performing graph similarity matching between the context graph and a pre-constructed semantic memory network; performing inference based on the matching results to generate a hierarchical task tree; executing atomic action sequences in the hierarchical task tree; and adjusting the action parameters of the atomic actions and the task execution path of the hierarchical task tree according to the semantic state change sequences of the dynamic entities and the context graph.

The invention relates to the technical field of three-dimensional point cloud data processing and registration, discloses a tunnel point cloud registration method based on slice contour map matching, and solves the problems of low registration precision, insufficient robustness and poor adaptability of an existing tunnel point cloud registration method in a scene with a low overlapping rate, strong noise and a repeated structure. According to the scheme, firstly, tunnel point cloud is preprocessed, noise points are removed, and coordinate scales are unified; and extracting the main axis direction of the tunnel through principal component analysis, and executing curvature-driven self-adaptive slicing along the main direction. For each slice, an Alpha Shape algorithm is used to extract a contour, a uniform point sequence is generated through polar angle sorting and resampling, and then a smooth curve is obtained through Catmull-Rom spline fitting. And then, the slice contour is converted into a graph structure, node matching and transformation estimation are carried out through graph similarity optimization, and high-precision registration of multi-site tunnel point clouds is realized.

The application provides a multi-source data fusion power knowledge processing method and device, medium and equipment, relates to the field of power data processing, and comprises the following steps: acquiring graph structure data corresponding to each preset data source of a target power system; performing function mapping on each node in each graph structure data to obtain standard function graph structure data corresponding to each preset data source; determining each target device node in different preset data sources according to each standard function graph structure data and a preset graph similarity algorithm; and fusing the graph structure data corresponding to each preset data source with each target device node as a fusion node to obtain a multi-source data fusion power knowledge graph. The multi-source data fusion power knowledge graph after fusion has the characteristics of data consistency, semantic unification and topological completeness.

The invention provides an irrigation area water distribution method based on topology and hierarchical perception graph similarity calculation, and relates to the field of intelligent water distribution, the method comprises the following steps: respectively modeling a historical water distribution scheme and a current water demand sub-graph into directed unweighted graphs; multi-dimensional feature coding is carried out on each node, and node enhancement representation is obtained after fusion; performing multi-scale feature extraction on the node enhanced representation by using a relational graph convolutional network, and introducing a gating mechanism to filter invalid node information to obtain final node embedding; on the basis of final node embedding, hierarchical constraint weighting attention aggregation double-path features are adopted, and graph-level representation is generated; calculating a similarity score between the two graphs, and performing weighted fusion to obtain a prediction graph editing distance; and recommending an optimal water distribution scheme from a historical scheme library according to the prediction map editing distance. According to the technical scheme, the matching task of the historical water distribution scheme graph and the current demand graph is converted into a scene adaptive graph similarity modeling problem, and a core basis is provided for intelligent recommendation of the irrigation area water distribution scheme.

The application discloses a graph similarity measurement method and system based on interactive attention, and belongs to the technical field of graph data processing and graph neural networks. The method comprises the following steps: obtaining a pair of images to be matched; performing node representation learning through a hybrid encoder of fused residual graph convolution and structure-aware self-attention to generate node embedding fusing local causal features and global semantic context; inputting the node embedding of the two graphs into a cross-graph interactive attention module to calculate a multi-view fine-grained node-level similarity matrix; adaptively fusing each view similarity matrix through similarity self-attention, and mining local topological matching patterns from the fused matrix through a convolutional neural network to finally map a scalar score representing the semantic similarity of the two graphs. The application converts the subgraph isomorphism problem into a learnable similarity measurement problem, can robustly identify the graph structure with structural variants and noise interference, and significantly improves the accuracy and efficiency of graph matching.

The invention discloses a multi-graph neural network integrated graph similarity evaluation method, which comprises the following steps of: constructing a graph similarity evaluation model comprising D graph neural networks, an embedded feature weighted fusion module, a graph cross mapping module and a graph similarity prediction module, the embedded feature weighted fusion module fuses the D embedded features of each graph, the graph cross mapping module performs cross mapping on the fused embedded features of the two graphs to obtain graph-level fusion representation features, and the graph similarity prediction module obtains a graph similarity prediction value according to the graph-level fusion representation features. And training the graph similarity evaluation model by adopting the training sample set, and evaluating the two input graphs by adopting the trained graph similarity evaluation model to obtain the graph similarity. According to the method, richer graph embedding is generated by using a plurality of graph neural networks, and graph-level fusion representation features are obtained by adopting graph cross mapping, so that the accuracy of graph similarity evaluation is improved.

There are provided systems and methods for pairwise graph querying, merging, and computing for account linking. A service provider may provide an account graph system to identify pairwise similarities between different accounts based on shared data that may be identified through one or more linking characteristics. When providing pairwise graph similarities, a service provider may receive a query identifying two or more accounts and / or an account with a parameter for graph exploration and querying.The service provider may utilize connection, link, or relationship graphs, queried and generated using a graph database, to determine pairwise similarities between the designated seed account and one or more selected accounts. The graph may include vertices for different queried data points and edges connecting such queries, where directionality of the edges or other vectors may be used to identify links or hops between accounts for data querying and exploration.

The application relates to the technical field of cheminformatics, and discloses a resistance distance-Kirchhoff iteration-based organic molecule similarity judgment method, which comprises the following steps: obtaining a simplified molecular linear input specification of a molecule to be compared, and converting the molecule to a weighted adjacency matrix representing the relationship between atoms and chemical bonds; expanding the matrix dimension and introducing a symbol parameter to construct an extended adjacency matrix, and converting a discrete connection structure into a parameterized continuous model; subsequently, the molecule graph is equivalent to a resistance network to perform Kirchhoff iteration, the pseudo-inverse of a Laplacian matrix is calculated, and a resistance distance matrix sequence representing global topology is obtained; a characteristic function in the sequence is extracted, integral operation is performed in a specified interval, and a quantitative graph similarity distance is output. The application effectively avoids the non-deterministic polynomial calculation bottleneck of traditional graph matching, completely retains the global topological information of the molecule, eliminates the isomorphism graph judgment ambiguity, and improves the discrimination degree and calculation efficiency of the isomer molecule.

The application discloses a deep learning-based automatic label generation method for clothing categories, and belongs to the field of clothing digital management, comprising the following steps: defining a parameterized clothing pattern topology skeleton library covering target categories, each skeleton being a two-dimensional graph structure composed of key points and elastic constraint edges; fitting an input clothing image with the clothing pattern topology skeleton, matching the skeleton contour with the clothing contour in the image through deformation to obtain a skeleton instance and deformation parameters; performing graph similarity matching on the skeleton instance and the skeletons in the skeleton library to determine a matching category and output a deformation parameter vector; and deforming the skeleton of the matching category according to the deformation parameter vector. The application can realize the construction of a parameterized clothing pattern topology skeleton library, the execution of causal intervention calculation and dynamic arbitration, effectively solves the problem that existing deep learning methods are prone to confusing visually similar but semantically different clothing categories, and improves the accuracy of automatic label generation for clothing categories.

The invention belongs to the technical field of intelligent traffic system and deep learning crossing, and particularly relates to a high-precision traffic flow prediction method based on dynamic space-time modeling and multi-graph fusion, which comprises the following steps: initializing a road connection graph, a similarity graph and a self-adaptive dynamic graph; mapping the original features to a high-dimensional space for output; periodic features are extracted through an autocorrelation periodic modeling module, dynamic space features are captured through a diffusion diagram attention module, static and dynamic diagram feature fusion is achieved through a multi-diagram fusion module, and a fusion feature matrix is output; and constructing an encoder-decoder and cross attention mechanism, and generating a traffic flow prediction result. According to the method, through integrating four core mechanisms of dynamic modeling of diffusion diagram attention, gating fusion of three types of diagrams, adaptive dynamic pause optimization and FFT-based self-correlation periodic capture, the robustness of the model in abnormal scenes such as extreme weather and sudden accidents is improved, and reliable support is provided for an intelligent transportation scheduling system.