307 results about "Graph embedding" patented technology

The invention discloses a multi-modal knowledge graph construction method, and relates to the knowledge engineering technology in the field of big data. The method is realized through the following technical scheme: firstly, extracting multi-modal data semantic features based on a multi-modal data feature representation model, constructing a pre-training model-based data feature extraction model for texts, images, audios, videos and the like, and respectively finishing single-modal data semantic feature extraction; secondly, projecting different types of data into the same vector space for representation on the basis of unsupervised graph, attribute graph, heterogeneous graph embedding and other modes, so as to realize cross-modal multi-modal knowledge representation; on the basis of the above work, two maps needing to be fused and aligned are converted into vector representation forms respectively, then based on the obtained multi-modal knowledge representation, the mapping relation of entity pairs between knowledge maps is learned according to priori alignment data, multi-modal knowledge fusion disambiguation is completed, decoding and mapping to corresponding nodes in the knowledge maps are completed, and a fused new atlas, entities and attributes thereof are generated.

Approaches to embedding source graphs into targets graphs in a computing system are disclosed. Such may be advantageously facilitate computation with computing systems that employ one or more analog processors, for example one or more quantum processors.

This invention relates to supervised or unsupervised classification of biological datasets. Specifically, the invention relates to the use of Graph Embedding as a method of reducing dimensionality thereby improving supervised classification of classes, both conventional and new ones.

The invention discloses a network graph embedding method based on edges. The method comprises steps that an Edge2vec algorithm model is constructed; a random gradient descent method is utilized to train the Edge2vec algorithm model; the Edge2vec algorithm model is utilized to realize network graph embedding. The method is advantaged in that the edges of the network graph are directly mapped to thelow dimension vector space through a neural network model based on an automatic depth encoder, the local neighbor degree information and the global neighbor degree information between the edges can be reserved, compared with the prior art, the structural information of the edges of the network graph can be effectively reserved, better performance can be realized for network graph analysis tasks of the edges, and the method can be applied to the network graph analysis tasks of the edges.

A method for detecting malicious program behavior includes performing program verification based on system activity data, analyzing unverified program data identified from the program verification to detect abnormal events, including analyzing host-level events to detect abnormal host-level events by learning a program representation as a graph embedding through an attentional architecture based on an invariant graph between different system entities, generating detection results based on the analysis, and performing at least one corrective action based on the detection results.

The invention discloses a graph embedding low-rank sparse representation recovery sparse representation face recognition method and belongs to the technical field of computer vision and mode recognition. The method comprises the steps that firstly, a graph embedding low-rank sparse representation recovery method is provided, a clean training sample data matrix with the high discrimination power can be recovered from a training sample data matrix, and meanwhile a training sample data error matrix is obtained; secondly, the sparse representation coefficient of face data to be recognized is obtained by using the clean training sample data matrix as a dictionary, using the training sample data error matrix as an error dictionary, and adopting the norm optimization technology; thirdly, according to the sparse representation coefficient of the face data to be recognized, class association reconstruction is carried out on the face data to be recognized; finally, a face image to be recognized is recognized based on the class association reconstruction error of the face data to be recognized. According to the graph embedding low-rank sparse representation recovery sparse representation face recognition method, the problem that the face recognition effect is poor on the conditions that a training sample image and an image to be recognized are polluted by noise or partially blocked can be solved.

Approaches to embedding source graphs into targets graphs in a computing system are disclosed. Such may be advantageously facilitate computation with computing systems that employ one or more analog processors, for example one or more quantum processors.

The invention relates to an enterprise operation risk prediction method. The method includes collecting the information data inside and outside an enterprise, and performing the data fusion processingon the collected information data; constructing and drawing a knowledge graph by utilizing the processed information data; extracting operation risk characteristic information data of an enterprise needing risk prediction from the knowledge graph by using a graph embedding method, and constructing an operation risk model by using the operation risk characteristic information data; and inputting the enterprise information data of which the risk needs to be predicted into the operation risk model, so that the development risk of the enterprise can be predicted and judged, correct decision on risk management of the enterprise of which the risk needs to be predicted is facilitated, and the asset safety of the enterprise of which the risk needs to be predicted is protected.

The invention discloses a knowledge reasoning method based on a multi-modal knowledge graph, and aims to enable knowledge reasoning reliability and accuracy to be higher and enable the knowledge reasoning method to have stronger modeling and reasoning capabilities. The method is realized through the following technical scheme: different information is fused based on multi-hop reasoning of a large-scale knowledge base; attribute completion is performed on the attribute missing graph through attribute graph embedding, structured information is extracted from unstructured and semi-structured documents or sentences, and a dynamic heterogeneous graph embedding model is constructed for multi-type characteristics of the multi-modal knowledge graph through heterogeneous graph embedding; feature learning of semi-structured knowledge, structured knowledge and different types of non-structured knowledge is achieved, and multi-modal knowledge graph features are obtained and serve as input for knowledge reasoning based on a graph neural network GNN; an inference path is generated, and a plurality of types of inference paths are constructed; and classification, edge prediction and frequent subgraphs of node types are calculated on the graph, a knowledge reasoning task is generated, and multi-step complex knowledge reasoning is completed.

The invention discloses a sequence recommendation method and device and a computer readable storage medium. The method comprises the steps of combining a bidirectional graph formed by a user set-project set with a knowledge graph in advance and unifying the bidirectional graph and the knowledge graph into a mixed knowledge graph; inputting a historical interaction sequence of the to-be-recommendeduser and the mixed knowledge graph into a sequence recommendation model; the model comprises a knowledge graph embedding module, a graph attention network and a recurrent neural network. A knowledgegraph embedding module encodes all nodes of the mixed knowledge graph into vectors, and a graph attention network recursively updates the embedding of each node according to the embedding of each nodeand the embedding of adjacent nodes so as to capture a global user-project relationship and a project-project relationship; the recurrent neural network encodes the user interaction sequence items toobtain dynamic preferences of the user; and finally, determining recommendation sequence information of the to-be-recommended user according to the output of the model, thereby performing high-accuracy sequence recommendation based on a high-order dependency relationship between entities in the knowledge graph and local graph contexts.

The invention discloses a congestion index prediction method combining a road network topological structure and semantic association. The method comprises the following steps: (1) establishing an undirected graph based on a space topological structure of a road network; (2) firstly calculating the similarity between the historical congestion index data of the road, then establishing a weighted undirected graph based on the similarity, and finally embedding the weighted undirected graph to obtain a semantic vector for representing the road; And (3) extracting short-term congestion index changecharacteristics on the basis of the graph convolutional network, extracting long-term congestion index change characteristics on the basis of the recurrent neural network, and fusing road semantic vectors on the basis to establish a prediction model. According to the method, spatial topology association and historical semantic association of the road network are considered at the same time, and the prediction capability of the model is improved; A graph convolutional network is adopted to model a road network topological structure, and graph embedding is adopted to model road network semanticassociation, so that the road network topological structure and the semantic association can be processed by a deep neural network.

The embodiment of the invention provides a behavior prediction model training method. The method comprises the steps: firstly determining a plurality of sample users for a target object, enabling anyfirst sample user to correspond to a first sample hard tag, and indicating whether a specific behavior is carried out on the target object or not; performing graph embedding processing on the user-object bipartite graph based on the trained graph neural network to determine an embedding vector set; determining a sample user feature vector corresponding to the first sample user and a target objectfeature vector corresponding to the target object, determining a specific behavior probability that the first sample user makes a specific behavior to the target object as a first sample soft label, and inputting the sample user feature vector into a first behavior prediction model to obtain a behavior prediction result; and training the first behavior prediction model by using a first loss item determined based on the prediction result and the first sample hard tag and a second loss item determined based on the prediction result and the first sample soft tag.

The invention discloses a collaborative filtering recommendation algorithm based on a graph convolution attention mechanism. The method comprises the steps of firstly collecting and processing data and dividing a data set, secondly constructing a GACF model, and finally training the model and recommending by predicting an association score between a user and an item. According to the graph convolution attention mechanism collaborative filtering model provided by the invention, firstly, user-project interaction information is mapped to a vector space by using a graph embedding technology, then, the embedding expression of a user-project interaction graph is learned through a graph convolution network, and then, different weights are allocated to neighbor nodes by using an attention mechanism. By aggregating the feature information of the neighbor nodes, the weight between the neighbor nodes can only depend on the feature expression between the nodes, so that the generalization ability of the model is improved, and finally, a plurality of embedded vectors learned by a graph convolution layer are weighted and aggregated to obtain the association score between the user and the project.

The embodiment of the invention provides a heterogeneous graph embedding-based network space security threat detection method and system. The method comprises the steps of obtaining entity behavior data; associating all data items in the entity behavior data according to the meta-attribute association relationship to obtain a data item sequence, and constructing a heterogeneous graph based on thedata item sequence; converting each node in the heterogeneous graph into a low-dimensional vector based on a graph embedding learning method, and obtaining vectorized expression of each node; and analyzing and processing the features of the vectorized expression to judge whether the data item corresponding to the vectorized expression is a malicious behavior or not. According to the embodiment, the heterogeneous graph for threat detection is established, entity behavior data items are simplified and represented in a vectorized mode, data item level threat detection for network space security is provided, later manual correction is not needed, labeled data items serve as training samples, and the detection precision and the detection feasibility are effectively improved.

The embodiment of the invention provides a method and a device for embedding a relation network graph in a multi-dimensional space through a distributed architecture. The distributed architecture comprises a plurality of processing devices, wherein one set of nodes of the relation network graph is distributed to each processing device. The method is executed by a certain processing device, and comprises the steps of: determining first-order adjacent nodes of a first node for which the processing device is responsible, determining processing devices corresponding to the first-order adjacent nodes; determining an initial embedding vector of the first node as a current order embedding vector; employing the current order embedding vector of the first-order adjacent nodes to calculate the nextorder embedding vector of the first node, and employing the next order embedding vector to update the current order embedding vector; and continuously executing the updating step until the current order quantity reaches a preset search depth, and taking the current order embedding vector at present as an embedding vector of the multi-dimensional space. Therefore, the distributed graph embedding method and device can effectively embed the relation network graph into the multi-dimensional space.

The invention belongs to the technical field of a spatial database, specifically relates to a method for finding an optimal path in a road network based on graph embedding. The method comprises the following steps of: defining point sets M1, M2, ..., and Mn of n attributions in a road network; and defining a starting point s and an ending point t; providing upper and lower limits according to a graph embedding frame by taking s as the starting point; finding out a greedy path Rg by finding the nearest neighbour with iteration; then traversing the point sets M1, M2, ..., and Mn of the n attributions according to a sequence; pruning the existing path by combining with the upper and lower limits and taking the length of the greedy path as the upper limit; and finally counting the residual candidate paths precisely to find the optimal path. The finding method based on graph embedding provided by the invention has the advantages that the searching space is greatly reduced, and the finding efficiency is high.

The invention discloses a tensorized embedded hyperspectral image classification method based on a sparse low-rank regular graph. The method overcomes the shortcomings in the prior art that classification is performed without fully utilizing hyperspectral image sample information and neighborhood information. The method comprises steps of: (1) inputting a hyperspectral image; (2) determining a training sample set and a test sample set; (3) constructing an adjacent matrix of the training sample set; (4) constructing a sparse low-rank regular graph; (5) determining an airspace training sample set and an airspace test sample set; (6) constructing a tensorized graph embedding model; (7) performing dimension reduction on the airspace test sample set; (8) performing hyperspectral image classification; and (9) outputting classified images. The method has an accurate classification effect at the edges and the homogeneous regions of the hyperspectral images and can be used for classifying the hyperspectral images.

The invention discloses a multi-visual angle gait recognition method and a system based on higher-order tensor subspace learning, which belong to the field of intelligent recognition. A gait video is acquired from multiple representational angles, and a gait sequence image is obtained through framing interception; background extraction, background subtraction and binary processing are carried out on the gait sequence image respectively, black and white visual effects are presented, and a contour sequence under the multiple visual angles is obtained; the contour sequence is converted to tensor data; a higher-order discriminant tensor subspace analysis algorithm based on graph embedding obtained after expanding DTSA on the basis of multilinear discriminant analysis and a graph embedding principle is used for carrying out dimension reduction and feature extraction on the tensor data; and according to the extracted and obtained multi-visual angle gait features, the gait features are subjected to similarity measurement, and a recognition result is obtained. The method is simple, the cost is low, person identity authority detection and disguised person identity authentication can be automatically carried out on a particular place, and safety protection on the monitored place and identity authentication in multiple conditions can be effectively improved.

The invention provides a cargo source route similarity calculation method, system and device and a storage medium. The method comprises the steps of collecting multiple pieces of cargo source route data within a set time range; Constructing a route map structure model according to the cargo source route data; Sampling the route map structure model to obtain data of a cargo source route serving asa sample; Inputting the data of the sample into a graph embedding model to obtain an embedding vector of each cargo source route; And calculating the similarity between the two cargo source routes byadopting the embedded vectors of the two cargo source routes. The scheme of the invention is adopted to calculate the similarity of different cargo source routes based on a graph embedding technology.On the basis of historical operation data of a plurality of drivers within a period of time of the platform, the similarity of all the cargo source routes is comprehensively calculated according to the selected times of different cargo source routes, the execution time and the identities of the execution drivers, and the correlation degree of all the cargo source routes can be reflected more accurately.

The invention discloses a domain knowledge pushing method based on a knowledge graph. The domain knowledge pushing method comprises the steps of collecting domain knowledge texts to construct a text knowledge base; performing semantic analysis and topic modeling on the knowledge base text; obtaining semantic distribution vectors of the nodes through domain knowledge graph embedding; establishing task context features according to the user task description and the task theme; performing entity alignment according to domain entities in the task description and the knowledge graph, performing feature extension based on graph node paths and graph node semantic distribution features, and performing task associated knowledge recall; performing text similarity calculation on the recalled text and the user task text to obtain candidate text scores; and pushing the sorting result text to the user according to the score. The text matching degree and the user experience of domain knowledge pushing are improved through node associated knowledge of the knowledge graph and a graph embedding technology.

The invention discloses a knowledge graph completion method and device, a storage medium and electronic equipment, and belongs to the technical field of computers. The knowledge graph completion method comprises the steps of obtaining a to-be-verified target knowledge text, generating a plurality of triples according to the target knowledge text and a preset knowledge graph, calculating each triple to obtain a corresponding confidence coefficient, verifying a target triple based on the corresponding confidence coefficients, and complementing the knowledge graph according to the verification result. Therefore, according to the method, the mixed model combining the text coding technology and the graph embedding technology is provided to learn the context and the structured knowledge at the same time, the reliable triple confidence score is obtained, advantage complementation of the two methods is achieved, the calculation overhead is remarkably reduced, and the complementation accuracy is improved. The invention further provides an adaptive integration scheme, scores of the coding method and the graph embedding method are fused in an adaptive mode, and the knowledge graph completion accuracy is further improved.

The invention discloses a heterogeneous graph embedding learning method based on an attention mechanism. The heterogeneous graph embedding learning method comprises the following steps: step 1, converting all nodes in a heterogeneous graph into a unified feature space through a type conversion matrix; 2, designing attention weights of a type-level attention learning given node for different types of neighbors; step 3, designing a neighbor node attention weight of node-level attention learning based on a meta-path, and performing weighted aggregation according to the attention weight to obtain node embedding based on a specific meta-path; 4, designing attention weights of different meta-paths of semantic-level attention learning, and performing weighted aggregation on node embedding based on different meta-paths according to the attention weights to obtain final node embedding; step 5, performing prediction training on the node labels; and step 6, designing a loss function, and performing model optimization training by using a back propagation algorithm. The work of the invention provides a new research thought for how to apply the attention mechanism to the heterogeneous graph.

The invention discloses a cross-language information retrieval method based on a conceptual graph. The concept map-based cross-language information retrieval method is a cross-language similarity evaluation framework CG-CLIR method, which comprises the following steps: S1, based on the concept edge representation of Skip-Gram; S2, high-order semantic embedding and similarity calculation of the fusion side information; the step S1 includes text preprocessing, concept graph edge semantic embedding. The step S2 includes LSTM based graph level semantic embedding, graph embedding based similarity calculation. According to the method, in text cross-language information retrieval, language obstacles are crossed, and the semantic retrieval effect is achieved under the condition that translation isnot conducted.

The embodiment of the invention provides a training method of a graph neural network. The method comprises the following steps: firstly, obtaining a relational network diagram, wherein the relational network diagram comprises a plurality of object nodes corresponding to a plurality of business objects; next, for each object node, fusing the node feature of the object node with the node feature of the neighbor node of the object node to obtain a fusion feature of the object node, and forming a fusion feature matrix by a plurality of fusion features corresponding to the plurality of object nodes; utilizing the graph neural network to perform graph embedding processing on the relation network graph to obtain a plurality of embedded vectors corresponding to the plurality of object nodes, the graph neural network comprising an activation function, and determining a plurality of prediction results based on the plurality of embedded vectors; determining a product matrix before and after the fusion feature matrix is processed by the activation function; and determining a training gradient of parameters in the graph neural network based on the product matrix, the plurality of prediction results and the service labels, and further updating the parameters in the graph neural network based on the training gradient.

The invention discloses a mobile advertisement fraud detection method based on heterogeneous graph embedding. The method comprises the following steps of 1) acquiring the mobile advertisement log dataand preprocessing the data; 2) extracting the association relationship data of a user, an application and an advertisement, and constructing a weighted heterogeneous graph; 3) defining a meta-path, setting the walking frequency and the longest step length of each node, traversing the weighted heterogeneous graph nodes, and constructing a node meta-path random walking sequence; 4) constructing low-dimensional space dense vector representation of nodes in the weighted heterogeneous graph by using a language model; 5) defining a label to form tested data; 6) constructing a mobile advertisement fraud detection model; 7) inputting the mobile application tested data of the training part into a mobile advertisement fraud detection model for training to obtain a mobile advertisement fraud detection model, and 8) carrying out fraud detection on the mobile application by adopting the mobile advertisement fraud detection model, thereby effectively detecting the fraud mobile application by utilizing the entity association relationship in the mobile advertisement system.

The invention discloses an Ethereum phishing fraud detection method and device based on graph classification, and the method comprises the steps: extracting a target node and preset-order neighbor nodes from an Ethereum network, and enabling the preset-order neighbor nodes to comprise a first-order neighbor node and a second-order neighbor node; constructing a second-order transaction sub-graph network taking the target node as a center node according to the first-order neighbor node and the second-order neighbor node; refining the second-order transaction sub-graph network according to the related transaction information data of each node in the second-order transaction sub-graph network to obtain a target transaction sub-graph network; extracting features in the target transaction sub-graph network by adopting a preset graph embedding algorithm to obtain a network representation vector; and inputting the network representation vector into a preset classifier for binary classificationprocessing to obtain a target fishing node. The technical problems that in the prior art, manual feature limitation is obvious, the processing process is complex, and consequently time and operationcost is high can be solved.

A method for extracting information from a table includes steps as follows. Characters of a table are extracted. The characters are merged into n-gram characters. The n-gram characters are merged into words and text lines through a two-stage GNN mode. The two-stage GNN mode comprises sub steps as: spatial features, semantic features, CNN image features are extracted from a target source; a first GNN stage is processed to output graph embedding spatial features from the spatial features; and a second GNN stage is processed to output graph embedding semantic features and graph embedding CNN image features from the semantic features and the CNN image features, respectively. The text lines are merged into cells. The cells are grouped into rows, columns, and key based on one or more adjacency matrices, a row relationship among the cells, a column relationship among the cells, and a key-value relationship among the cells.

A drug recognition method based on biomedical knowledge map reasoning comprises the following steps of S1, downloading the biomedical text data; S2 constructing a biomedical knowledge map; S3 constructing a drugs-target-disease relationship data set; S4 using a graph embedding method to learn graph representation; 5 training a drug discovery model based on a long-short memory neural network; S6 performing a drug identification step by using the trained model. The method of the invention is applicable to searching for potential therapeutic drugs for diseases, and is not limited to diseases andtypes of drugs, and has great significance to discover drugs that can be used to treat diseases from the literature.