23 results about "Network representation learning" patented technology

The present invention discloses a drug-target prediction method based on a graph-attributed neural network, comprising the following steps: S1, constructing a multi-source heterogeneous biological network and uniquely identifying drugs, proteins, and diseases; S2, calculating the similarity between any two diseases based on the disease module theory of the human protein-protein interaction network; S3, using each biological entity pair and the corresponding similarity value as a training sample for the graph attention neural network representation learning phase; S4, using the training samples to drive the graph attention neural network learning to obtain a representation vector for each entity; and S5, using the trained drug-target prediction model to predict drug-target interactions. This invention reduces the dependence of deep learning models for drug-target interaction prediction on training samples, thereby improving prediction performance.

The invention provides an Internet infrastructure knowledge graph construction method based on network representation learning and classification and grading thought. The method comprises the following steps: constructing an ontology model of Internet infrastructure; collecting map data; utilizing the ontology model and the atlas data to create entities and extract relationships between the entities; classifying and grading the entities to obtain classification results and grading results corresponding to the entities; an initial knowledge graph is constructed through the relations between the entities, and each entity in the initial knowledge graph is additionally associated with the corresponding classification result and grading result; and based on the network representation learning and the graph data, updating the edges, the classification result and the grading result in the initial knowledge graph to obtain a final knowledge graph. The knowledge graph is constructed and updated through network representation learning and classification and grading thoughts, so that the dynamic association relationship between the entities is accurately captured, wrong association edges are screened out, and the network representation accuracy and availability of the entities are improved.

Disclosed herein are a method and apparatus for analyzing a brain-inspired neural network, the method being performed by an apparatus for analyzing the brain-inspired neural network, the method including converting an input brain-inspired neural network into a computational graph, performing attention computation on the computational graph based on a graph attention network, and outputting a result of network representation learning for the brain-inspired neural network based on the attention computation.

The invention relates to the technical field of machine learning, in particular to a social network anomaly detection system and method based on network representation learning, and the system comprises a heterogeneous feature extraction layer, a cross-modal fusion layer, an anomaly decision-making layer, a joint optimization layer and a fairness module. The method comprises the following steps: processing text data obtained based on a social network to obtain network structure data and semantic behavior data; structural features are extracted through a GCN-M model by the heterogeneous feature extraction layer, semantic features are extracted through a dynamic fine tuning BERT model, and a cross-modal fusion layer fuses the structural features and the semantic features by adopting a dual-channel dynamic attention mechanism to obtain a fusion feature matrix and sends the fusion feature matrix to an anomaly decision-making layer; according to the method, the feature extraction capability and generalization performance of the model are enhanced, deep fusion of features is realized, and a user behavior mode can be described more comprehensively, so that the accuracy of anomaly detection is remarkably improved.

The present invention provides a hierarchically structured, fast heterogeneous information network representation learning method and device, comprising first acquiring the data to be mined, performing data cleaning, processing, and encoding, and constructing a heterogeneous information network; then, based on the heterogeneity and topological structure of the network, iteratively reducing the network using a heterogeneous reduction network model to obtain a series of hierarchical heterogeneous information networks of decreasing scale; then, learning the node representation of the smallest heterogeneous information network, and using this node representation to train a heterogeneous refinement model; finally, iteratively refining the network based on the mapping relationship between the heterogeneous refinement model and the hierarchical network to obtain the node representation of the constructed heterogeneous information network. The present invention can reduce the scale of large-scale heterogeneous information networks through reduction technology, thereby reducing the time and space complexity of the representation problem and rapidly learning the node representation in the network.

The application relates to the technical field of machine learning, in particular to a social network anomaly detection system and method based on network representation learning, which comprises a heterogeneous feature extraction layer, a cross-modal fusion layer, an anomaly decision layer, a joint optimization layer and a fairness module; text data obtained based on a social network is processed to obtain network structure data and semantic behavior data; the heterogeneous feature extraction layer is used to perform structure feature extraction through a GCN-M model, a BERT model is dynamically fine-tuned to extract semantic features, the cross-modal fusion layer adopts a double-channel dynamic attention mechanism to fuse the structure features and the semantic features to obtain a fusion feature matrix and send the fusion feature matrix to the anomaly decision layer; the anomaly decision layer realizes anomaly probability output through a deep forest classifier; the application enhances the feature extraction capability and the generalization performance of the model, realizes deep fusion of features, can more comprehensively depict user behavior patterns, and significantly improves the accuracy of anomaly detection.

The present application relates to a dynamic network representation learning method for node attribute preservation. The method includes: obtaining a time-dependent social network sample; constructing an inductive graph convolution model; determining the sampling order of neighbor nodes of a target node in the time-dependent social network sample according to the number of aggregate convolution modules of the inductive graph convolution model, and biased sampling from the highest sampling order downward layer by layer according to the order of interaction time between the neighbor nodes and the target node to obtain a sampling node set and a neighbor sequence queue for each layer; through #imgabs0# aggregate convolution modules, according to each layer's sampling node set and neighbor sequence queue, aggregate the neighbor node attribute vectors of the corresponding order layer by layer, and use the output of the last aggregate convolution module as the embedding vector of the target node; training the inductive graph convolution model, and using the trained inductive graph convolution model to perform representation learning to complete the social network analysis task. This method can learn more accurate node representations.

The present invention relates to the field of e-commerce network data mining and discloses a method and system for discovering e-commerce network communities based on semi-supervised network representation learning. The method comprises steps of constructing an e-commerce network adjacency matrix, a feature matrix, and a partial user label matrix; constructing and optimizing an e-commerce network community discovery model based on the semi-supervised network representation learning; and outputting the results of the e-commerce network community discovery model based on the semi-supervised network representation learning. The e-commerce network community discovery model based on the semi-supervised network representation learning is constructed using two independent GCN encoders, and a joint optimization objective function is designed, and the model is trained by minimizing the objective function. The system comprises a computer processor and memory, an e-commerce network adjacency matrix, a feature matrix, and a label matrix unit, a training unit for the e-commerce network community discovery model based on the semi-supervised network representation learning, and a unit for outputting the results of the e-commerce network community discovery model based on the semi-supervised network representation learning.

The present disclosure proposes a network representation learning method across medical data sources, comprising: S1, generating medical network data comprising a source network and a target network; S2, randomly sampling a set number of nodes from the source network and the target network; S3, obtaining an L-layer neural network, and calculating the structural features and expression features of the source network and the target network respectively for each layer, and calculating the distance loss between the network features of the source network and the target network; S4, obtaining the output of the source network in the L-layer neural network, and calculating the loss value according to the classification loss and the distance loss, and updating the parameters of the algorithm according to the back propagation algorithm; S5, repeating steps S2-S4 until the entire algorithm converges, so that the accuracy of the algorithm for disease classification no longer rises within multiple iterations. The present disclosure considers the problem of inconsistent data distribution between different hospital data sources, and through the extraction of network structural information and node attribute information and the minimization of feature distance, the information loss is compensated, which has a wide application space.

The application discloses a point of interest recommendation method based on meta-path enhanced view contrast learning, and the method comprises the following steps: constructing an initial position social network and generating node initial features; determining and generating corresponding enhanced edges according to whether the geographical distance between node pairs based on different correlation relationships meets the corresponding preset threshold, so as to obtain an enhanced position social network; constructing a network mode view, a meta-path view and a meta-path enhanced view; performing meta-path level view contrast and graph level view contrast, and respectively calculating corresponding contrast losses; calculating the total loss of model training by linearly weighting the meta-path level contrast loss and the graph level contrast loss; and generating a user point of interest recommendation after optimizing model parameters. The application realizes the construction of enhanced edges by integrating time and space information to alleviate data sparsity, and improves the network representation learning quality and the point of interest recommendation accuracy by means of multi-level feature aggregation and multi-level view contrast.

The present invention discloses a method for learning representation of a dynamic social network, which relates to the field of network representation learning and solves the technical problem that the current method cannot retain the low-dimensional feature vectors of the evolution pattern and community structure of a real social network during learning, including the analysis of the triadic closure pattern and the community structure. After analyzing the triadic closure pattern, the present invention jointly models the evolution process of nodes and edges in the social network by combining important factors of the social network such as the influence of nodes, the similarity of nodes, and the community structure. This method can retain both the network structure of nodes and the dynamic features to learn the low-dimensional representation of nodes, and by capturing the evolution patterns of different nodes in the social network, the learned low-dimensional representation is more discriminative. Community is one of the most important features of a real social network, and the learned network representation can well reflect the structure of the community, which can help users obtain more useful information and better optimize the triadic closure process.

The application discloses a multi-center depression recognition method and system based on decoupling of cross-subject relationship networks, relates to the technical field of medical image processing and artificial intelligence diagnosis, and comprises the following steps: acquiring subject data of a plurality of collection centers, wherein the subject data comprises brain image data; constructing an initial individual brain function network based on the brain image data; performing individual brain network representation learning on the initial individual brain function network to obtain a subject-level brain network representation; and performing decoupling learning and joint optimization based on disease-related and collection center-related cross-subject relationship networks, and then guiding iterative updating of the individual brain network through a group-level disease discrimination representation, so that the common characteristics related to depression can be better mined, the influence of the differences between the collection centers on feature learning is weakened, the representation and discrimination ability of the model for disease characteristics is improved, the model has more stable performance in a multi-center scene, and the generalization performance of cross-center data is improved.

The invention relates to a dynamic network representation learning method oriented to node attribute storage. The method comprises the following steps: acquiring a time-dependent social network sample; constructing an inductive graph convolution model; determining a neighbor node sampling order of a target node in the time-dependent social network sample according to an aggregation convolution module number of the inductive graph convolution model, and starting layer-by-layer downward biased sampling from the highest order of the sampling order according to a sequence of interaction time of the neighbor node and the target node, obtaining a sampling node set and a neighbor sequence queue of each layer; through # imgabs0 # aggregation convolution modules, aggregating neighbor node attribute vectors of corresponding orders layer by layer according to a sampling node set of each layer and a neighbor sequence queue, and taking the output of the last aggregation convolution module as an embedded vector of a target node; and training the inductive graph convolution model, and performing representation learning by using the trained inductive graph convolution model to complete a social network analysis task. By adopting the method, more accurate node characterization can be learned.

The application discloses a multi-head attention mechanism-based super network representation learning method applied to breeding, relates to the technical field of crop breeding, and comprises the following steps: constructing a crop heterogeneous super network, and constructing an initial embedding matrix corresponding to all nodes; obtaining a normalized node representation vector corresponding to each tuple and a similarity score of each tuple; calculating the pairwise similarity between each two nodes; constructing a joint loss function, and optimizing the parameters of a target model; generating a final representation vector of a node in the crop heterogeneous super network by using the target model after parameter optimization; and predicting a crop variety according to the final representation vector. The application avoids complex graph conversion and explicit high-order combination calculation, improves the calculation efficiency and scalability, can more comprehensively and accurately encode semantic and structural information of phenotypic traits, and thus provides reliable and high-quality data basis for crop variety prediction based on the final representation vector.

The invention provides a hierarchical semantic perception-based network fraud identification method and system. The method comprises the following steps of: constructing a network transaction graph based on network transaction data; carrying out characterization representation on nodes and edges of the network transaction graph; a causal attention module is adopted to calculate causal association strength represented by the node features, and an attention weight is generated based on the causal association strength; carrying out importance measurement and sorting on neighbor nodes through attention weights to form a neighborhood set; performing hierarchical semantic feature extraction on the neighborhood set by adopting an attention aggregation information strategy; the hierarchical semantic features are input into a gating updating module, and dynamic weighted updating is carried out on information from a neighborhood and historical representation of nodes according to gating coefficients; a deep network representation learning structure is formed in the stacking hierarchical semantic extraction and gating updating process, and final feature representation of the nodes is generated; inputting the node feature representation into a classifier to obtain a prediction result; layered modeling is carried out for multi-source, heterogeneous and multi-layer semantic information in a network transaction relation graph, and interference of noise propagation on a detection result is reduced by explicitly describing risk contributions of different semantic levels and different relation types.

The invention discloses a teacher and bearing relationship prediction method for science and technology data fusion with a new subject atlas, and the method comprises the steps: carrying out the analysis and statistics of paper and patent data in a plurality of fields, constructing a network representation learning device through employing a network representation learning algorithm, and carrying out the prediction of the teacher and bearing relationship based on a deep neural network-mixed principal component analysis (PCA) algorithm and a pooling layer technology means. The method comprises the following steps: establishing a teacher-bearing relationship identifier, adding attribute network information, establishing a new subject map identifier based on subjects, performing classification optimization on different subjects, forming a data set of advisor-advisor advisor pairs and advised pairs by using the teacher-bearing relationship identifier, and combining the teacher-bearing relationship identifier and the new subject map identifier to obtain an advised-advisor advisor-advisor advisor-advisor advisor-advisor advisor map. A reliable machine learning prediction model is established to predict a tutor-student relationship, and statistics and modeling are carried out to calculate a cooperation network of the technology in different periods of talents in papers and patents, so that the teacher-bearing relationship of the talents is judged, for example, who is the tutor of the talents and which students are carried by the tutor of the talents, and other teacher-bearing relationships are judged.

The invention discloses a book borrowing trend prediction method and system based on big data analysis, and relates to the technical field of book management, and the method comprises the steps: obtaining a plurality of heterogeneous data sources, carrying out the vectorization of all unstructured data, and constructing a heterogeneous information network; carrying out network representation learning on the heterogeneous information network to obtain low-dimensional vector representations of all nodes, carrying out clustering recognition based on the low-dimensional vector representations, extracting potential user communities with similar hidden interest features, and generating corresponding group interest feature vectors; and for any potential user community, predicting the borrowing demand tendency of the potential user community to the books which are not stored, and generating book purchasing suggestions. The technical problem that the accuracy of book purchase and recommendation is insufficient due to the fact that an existing book borrowing prediction method cannot fully mine the potential interest demand of the user is solved, and the technical effects that the potential demand of the user is accurately recognized through big data analysis and heterogeneous data source fusion, and the accuracy of borrowing demand prediction is improved are achieved.

The present invention discloses a method and system for tracing the results of an artificial intelligence model based on expert rules, which is used to solve the problem of lack of explainability of the prediction results of the artificial intelligence model. The present invention first collects high-quality historical data of auto insurance cases from insurance companies, and converts the historical data of auto insurance cases into graph structure data according to expert rules; then constructs a graph network through the graph structure data, and obtains expert factor vectors and expert rule vectors based on network representation learning model training; then collects real-time data of auto insurance that are determined to be risk cases by the artificial intelligence model, calculates the set of triggered expert factors, and obtains the quasi-triggering rule vector by default filling; then calculates the similarity between the expert rule vector and the quasi-triggering rule vector; finally, by returning the quasi-triggering rules with high similarity, provides traceability for auto insurance experts. The method of the present invention ensures a low traceability time while also ensuring the quality of traceability.

The application discloses a kind of based on residual unit optimization's super-network representation learning method applied to breeding, it is related to crop breeding technical field, method includes: constructing crop heterogeneous super-network, obtaining the representation vector of each node, and obtaining at least one to be modeled high-order tuple, the node representation sequence corresponding to each to be modeled high-order tuple is input into target model, the similarity of each to be modeled high-order tuple is obtained;Based on the similarity of each to be modeled high-order tuple, the representation vector of each node in crop heterogeneous super-network is optimized;Using the representation vector of all nodes after optimization, the prediction of crop variety is carried out.The application solves the problem that the existing method is insufficient in super-edge fusion, thereby improving the depth and fidelity of learning representation from high-order structure information, and while ensuring the ability to model complex relationships, the operability and pertinence of the method in real breeding scenarios are improved.

The present invention relates to a method and device for detecting industrial equipment events based on hierarchical network representation learning. The method comprises: obtaining event data to be detected, obtaining attribute representations and mapping relationships through a pre-trained hierarchical network representation learning model, thereby obtaining an event representation corresponding to the event to be detected; calculating the Euclidean distance between the obtained event representation and the cluster centers of all normal and abnormal events, and classifying the event to be detected into the cluster with the smallest distance. Compared with existing technologies, the present invention has stronger anomaly detection capabilities and robustness.

The present disclosure provides a network representation learning method, device and equipment and a storage medium, wherein the network representation learning method utilizes a target neural network to perform network representation learning in a target field; the target neural network comprises N-level first neural networks and N-1-level second neural networks; the output of the t-level first neural network is the input of the t-level second neural network; the output of the (t-1)-level second neural network is the input of the t-level first neural network and the input of the t-level second neural network; the network representation learning method comprises: obtaining heterogeneous network data; inputting the heterogeneous network data into the target neural network to obtain a first network representation; determining training data based on the heterogeneous network data and the first network representation; training a logistic regression model and the target neural network based on the training data; and taking the second network representation output by the target neural network when the training of the logistic regression model and the target neural network is completed as a target network representation of a target field task.

The application provides a paper recommendation method and system based on hierarchical graph attribute network representation learning, and the method comprises the following steps: constructing a paper network based on a paper library; constructing a hierarchical graph through edge merging, structural equivalence merging, community merging and attribute merging based on the paper network; performing intra-layer network representation learning and inter-layer network representation learning on the hierarchical graph to generate paper node embedding representation, wherein the inter-layer network representation learning is performed through forward initialization and then backward initialization; and generating a paper recommendation list by calculating the similarity of embedding representation between paper nodes based on the paper node embedding representation, and recommending the papers in the paper recommendation list to users. The paper recommendation method based on hierarchical graph attribute network representation learning is adopted, so that the problem of low paper recommendation accuracy of the existing network representation learning method is solved from the root.

An inter-domain routing anomaly detection method based on network representation learning, comprises: performing network representation learning of Internet autonomous system relationships under an offline condition; monitoring a routing change of a plurality of inter-domain routing vantage points in real time, updating global routing in response to receiving a border gateway protocol update message, and recording a routing change when the routing change is detected; in response to detecting the routing change, calculating a path difference value before and after the routing change, and when the path difference value is greater than a threshold, determining that the routing change is an abnormal routing change; and performing attribution on the abnormal routing change, aggregating the abnormal routing change into an abnormal event, and issuing an alarm.