92 results about "Hypergraph" patented technology

The invention provides a storage resource optimization method and system based on a time sequence dependence hypergraph neural network, and belongs to the field of artificial intelligence computing. Based on medical health big data and computing resources, constructing and fusing data and computing resource dependency matrixes to obtain a static dependency relationship matrix; the method comprises the following steps: collecting a computing resource multi-source operation log, generating dynamic characteristics of each moment according to a fixed interval, introducing time sequence position coding and self-attention mechanism weighting in a sliding time window to obtain attention optimization characteristics, and generating a dynamic dependency weight matrix by combining a modeling historical hidden state and the dynamic characteristics; the static and dynamic dependency weight matrixes are fused to obtain a comprehensive dependency matrix, attention optimization features are used as nodes, hyperedges are constructed in combination with the comprehensive dependency matrix, and hypergraph association and other matrixes are generated; and inputting the matrix into a graph neural network, learning node representation in combination with a time sequence attention network, classifying nodes and mapping the nodes into scheduling actions, and realizing self-adaptive allocation of storage resources in combination with target function optimization of medical scene constraints.

The invention discloses an electricity consumption anomaly detection method based on a graph structure, and relates to the technical field of electricity consumption anomaly detection, and the method comprises the steps: obtaining historical electricity consumption data, environment variable data and power grid physical topology information of a power grid region, and carrying out the standardization processing of the historical electricity consumption data, the environment variable data and the power grid physical topology information, a standardized time sequence-environment data set is obtained; based on the standardized time sequence-environment data set, identifying a causal association relationship between electricity consumption and environment variables through a causal discovery algorithm, and constructing a causal perception heterogeneous graph containing region nodes and environment factor nodes; based on a causal perception heterogeneous graph, introducing a plurality of region nodes of which hyperedge connection is influenced by the same environmental event to form a dynamic hypergraph structure for representing a many-to-many environment-region coupling relationship; based on the dynamic hypergraph structure, a graph comparison learning task is constructed, positive and negative sample pairs are generated by applying disturbance to node features and the graph structure, and unsupervised pre-training is completed;

The invention discloses a multi-modal data drawing logical relationship analysis method, electronic equipment and a medium, and the method comprises the steps: generating a node set based on drawing image data and text data; generating a cross-modal hyperedge set based on the spatial proximity relationship, the visual feature similarity and the semantic correlation between the node sets; generating a hypergraph embedding input representation based on the node set and the cross-modal hyperedge set; the hypergraph is embedded into the input representation input improved hypergraph self-attention network model, and a hyperedge logic relation type and a corresponding hyperedge confidence coefficient are generated; generating a graph structure result based on the hyperedge logic relationship type and the node set, wherein the graph structure result meets the structure legality requirement; and performing hyper-parameter automatic adjustment and convergence control on the atlas structure result based on hyper-edge confidence, and generating an optimal atlas analysis model and a structured output result. According to the method, the reliability and the quality of analysis of component nodes, logic edge relationships and semantic structures in the drawing are improved.

The invention belongs to the field of product top-layer system design and modeling, and discloses an interactive feedback-oriented product top-layer system design method, which comprises the following steps of: constructing a multi-layer model based on a demand R, a function F, a behavior B and a structure S, establishing a knowledge graph for representing a semantic relationship between layers and between nodes on the same layer by using weighted directed edges, forming a layer-by-layer mapping matrix by using edge weights, and constructing a layer-by-layer mapping model; realizing weight decomposition from the demand to the structural unit; a constraint hypergraph is constructed on the node set, and structural unit combination constraints are depicted through mutual exclusion, dependency, collaboration and conflict hyperedges; physical parameters of the structural units are normalized into performance vectors, weighted aggregation is carried out according to weights, legality judgment and performance correction are carried out in combination with a hypergraph, a system-level performance evaluation result is obtained, candidate configuration schemes are compared and optimized according to the result, and a product configuration scheme meeting requirements and constraint conditions is formed. The method provides a model basis for subsequent product structure configuration optimization based on interactive feedback.

The invention discloses a highway node risk calculation method based on a graph attention model. The highway node risk calculation method comprises the following steps: collecting and preprocessing multi-source node feature data of highway nodes; constructing a road network topological graph; generating an input feature representation vector set by improving an input coding module of the HGAT model; obtaining a neighbor node set; a hypergraph attention calculation module of the HGAT model is improved, a topology centrality guiding attention mechanism is introduced, and a road node representation vector is generated; a road node risk representation vector is obtained by improving a deep fusion module of the HGAT model; a road node risk value is obtained through a risk prediction output module of the improved HGAT model; and outputting a highway node risk assessment result, thereby improving the precision and stability of highway node risk calculation.

The invention provides a method and a system for constructing an online rainfall forecasting model based on a space-time dynamic hypergraph neural network. The method comprises the following steps of: establishing a graph model of a hypergraph, an adjacent matrix of an original graph and an incidence matrix of the hypergraph; taking a hypergraph neural network mechanism as a core, and establishing an online rainfall forecasting model based on a space-time dynamic hypergraph neural network; splicing the time sequence and the space sequence of the rainfall data by adopting an encoder, inputting the spliced time sequence and space sequence into a multi-layer perceptron with a full-connection structure, and finally outputting a prediction sequence by adopting the multi-layer perceptron with the full-connection structure by a decoder; wherein the step of establishing the encoder comprises establishing a time embedding module for capturing meteorological elements and establishing a hypergraph attention network; wherein the step of establishing the decoder comprises the step of establishing a multi-layer sensor with a full-connection structure. And establishing an online learning mechanism, a seasonal sample storage pool and an online training mechanism based on mixed domain dynamic playback for correcting the output of the online rainfall forecasting model.

The present disclosure provides a Boolean formula unsatisfiability core prediction method based on hypergraph modeling. A CNF formula generated by a software and hardware formal verification problem coding in the field of electronic design automation is received from the outside, input into a hypergraph representation learning model, and modeling for solving a SAT problem corresponding to the software and hardware formal verification problem is obtained. The model is constructed in the following manner: first, SAT problem representation based on hypergraph is performed. Specifically, the CNF formula is modeled as a clause-literal hypergraph by using a SAT problem modeling method based on hypergraph and a message passing method based on hypergraph, and a clause correlation graph is further constructed to obtain problem representation. Then, a polarity-aware variable decomposition method is applied to the clause correlation graph, and the problem representation is structured and modeled. The hypergraph representation learning model is trained by using a training method based on polarity-aware consistency constraints, and is integrated with a SAT problem solver, so that the model can be used to solve the problems that the existing graph neural network-based SAT learning method has in the aspects of polarity modeling and high-order structure expression.

The invention discloses an event prediction method and system based on a time sequence hypergraph, and belongs to the technical field of event prediction. The method comprises the following steps: acquiring multi-source heterogeneous data of a target region, defining a unified time index, and generating a time sequence feature sequence of each variable of the region through preprocessing; then, identifying a variable causal relationship in the region based on a frequency domain anti-fact condition mutual information algorithm, and fusing the variable causal relationship with a time sequence evolution relationship to construct a time sequence hypergraph structure representing the interior of the region; a dynamic filter is used for filtering the graph, and deep features of the graph are learned by means of a multi-band spectrum gating mechanism, so that internal complex causal and time sequence modes are effectively captured; and finally, performing dichotomy prediction based on the learned graph representation, and outputting the occurrence probability of future events in the region. According to the method, accurate and explainable event prediction is realized, training and prediction do not need to cross regions, data privacy and calculation efficiency are guaranteed, and stronger robustness is shown for specific data distribution change of the regions.

The application discloses a hypergraph recommendation method based on multiple conversation awareness, uses a hypergraph to obtain high-order relationship representation between items, combines a co-occurrence graph and a local conversation graph to obtain item correlation representation within and between conversations, uses a gating mechanism to mine mutual beneficial information between multiple conversations, adaptively captures the dependency relationship of items in the conversation behavior sequence, and further provides more accurate recommendation results for users. The application uses the natural property of the hypergraph to model the complex multi-element relationship in the conversation, alleviates the hyperedge isomorphism problem by keeping the sequence information. The application establishes a co-occurrence graph structure and a local conversation graph structure, realizes the connection between similar user intentions in different conversations and potential behavior patterns in the same conversation, and can explore the mutual influence of complex cross-conversation and intra-conversation behavior patterns.

The application provides a recommendation method for optimizing multi-user representation based on a hypergraph motif, and relates to the fields of knowledge graphs, recommendation systems and the like. The method constructs a user-item hypergraph based on the historical interactions of the user, maps it to a user adjacency graph, and uses it as the initial representation of the user implicit relationship network. Then, the hypergraph motif is extracted to learn the interaction mode of the multi-user, the implicit relationship network is reconstructed, and the item features in the hypergraph and the user association features in the implicit relationship network are fused to obtain a high-order representation of the user. Finally, the preference information of the user is propagated in the knowledge graph, and the representation of the item is aggregated. The method uses the hypergraph motif to learn the interaction mode of the multi-user, establishes a user implicit relationship network with direction and weight, can learn the interaction information of the multi-user, and then fuses the implicit relationship features of the user and the historical interaction item features to optimize the representation of the user and improve the performance of the recommendation system.

This application discloses a hypergraph representation learning method and apparatus based on heterogeneity adaptive positional encoding, which can be applied to the field of graph neural network technology. This application constructs a heterogeneity adaptive positional encoding module and a heterogeneity-aware distinguishable aggregation module. A triple array is input into the heterogeneity adaptive positional encoding module to capture the first output representation corresponding to long-distance semantically related neighbor relationships in the hypergraph to be analyzed. The obtained triple array is then input into the heterogeneity-aware distinguishable aggregation module to obtain a second output representation that can distinguish the importance of different neighbor nodes in the hypergraph to feature aggregation. The first and second output representations are then fused to obtain the target node embedding representation. Based on the target node embedding representation, a pre-defined hypergraph neural network model is optimized using supervised learning, thereby achieving low-dimensional feature representation learning for heterogeneous hypergraph nodes and effectively improving the representation learning performance of heterogeneous hypergraphs.

Embodiments of the present application provide a point cloud matching method and device, and a storage medium. The method can include determining a consistency matrix between a first point cloud and a second point cloud, wherein an element in the consistency matrix represents a consistency size between a corresponding relationship of two points included in a first group of points and a corresponding relationship of two points included in a second group of points; constructing a hypergraph according to the consistency matrix between the first point cloud and the second point cloud; performing graph convolution processing on the hypergraph to obtain at least two groups of initial transformation matrices; and obtaining a target transformation matrix between the first point cloud and the second point cloud according to the at least two groups of initial transformation matrices. The high-order consistency represented by the hypergraph structure is more accurate, and is more robust to abnormal points and noise. The method can provide more accurate information for identifying and registering inlier points, and is helpful for effectively registering point clouds in a low overlap rate situation.

This invention belongs to the field of conversation recommendation technology, and specifically provides a conversation recommendation method based on multi-intent expression using cross-category heterogeneous hypergraphs. The method includes preprocessed data and the following steps: S1: The preprocessed data is input to a category recognition processing module, which uses a Latent Dirichlet Allocation (LDA) model to mine the latent category distribution of items and outputs category-classified items; S2: The preprocessed data and category-classified items are respectively input to a self-loop star graph module and a cross-category heterogeneous hypergraph module to construct the graph structure, outputting a self-loop star graph and a cross-category heterogeneous hypergraph. This invention combines the conversational intra-conversational structure modeling of self-loop star graphs with the semantic association expression of cross-category heterogeneous hypergraphs to construct a multi-intent modeling framework that can simultaneously capture local behavioral dependencies and global interest transfers. It dynamically identifies latent intent patterns and multi-category preference features even without long-term user history information, thereby significantly improving the accuracy, personalization, and robustness of conversation recommendations.

The invention discloses a hypergraph propagation source positioning method based on interaction enhancement. A traditional traceability method mainly depends on a pairwise interaction hypothesis of a simple graph structure, and the positioning precision under a complex topology is limited. The method comprises the following steps: firstly, constructing a hypergraph propagation model and obtaining a complete observation snapshot of a network node; then, extracting the dynamic state and global spectrum features of the nodes, and constructing a basic feature vector; an improved double-flow interaction hypergraph neural network is utilized, hyperedge internal key information is aggregated through local attention flow, and multi-scale topology diffusion flow is utilized in parallel to capture overall long-range dependence; and finally, introducing a gating residual fusion mechanism to adaptively integrate the double-flow features, and accurately outputting the propagation source probability based on a weighted cross entropy loss function. According to the method, the path reconstruction deviation of a traditional method can be effectively overcome and the robustness can be improved in the aspect of processing high-order interaction and long-distance dependence. According to the method, the performance bottleneck based on simple graph traceability is broken through, and powerful technical support is provided for propagation source positioning.

The present application relates to the technical field of multispectral image fusion, and discloses a real-time multispectral image fusion system based on deep learning. The system comprises a multispectral image atlas module, a flow atlas synthesis module, a personalized flow extraction module, a flow vertex labeling module and a fusion flow verification module. The multispectral image atlas module stores a plurality of image category templates of graph structures, each template corresponding to a fusion flow graph containing processing vertices and connecting edges; the flow atlas synthesis module fuses a plurality of templates into a unified hypergraph through vertex clustering and edge redirection; the personalized flow extraction module analyzes user demand text and converts it into a demand atlas, and matches a subgraph from the unified hypergraph as a basic fusion flow; the flow vertex labeling module identifies adjustable vertices and adds labels; and the fusion flow verification module verifies logic through a deep learning model, detects abnormal vertices and generates a report. The system is suitable for multiple scenarios and improves the flexibility and reliability of the fusion flow.

The invention discloses an analysis scheduling method and system based on multi-label routing. The method comprises the following steps: receiving multi-label hypergraph data, performing tuck decomposition on a hypergraph adjacency tensor, applying sparse constraint, extracting a core multi-label dependency structure, and generating a low-rank factor matrix; extracting hyperedge embedding vectors according to the low-rank factor matrix, compressing the hyperedge embedding vectors into binary hash codes through a locality sensitive hash module, and splicing the binary hash codes with the residual floating point features to form mixed hyperedge embedding vectors; for nodes in the graph neural network, attention weights are calculated through a multi-label compatibility function based on mixed embedding vectors, neighborhood aggregation is carried out, node states are updated through a gating circulation unit, and node embedding representation is obtained; and performing routing scheme generation on the node embedded representation through a path search algorithm, evaluating different schemes through a path scoring function, and outputting an optimal scheduling route according to a scoring result. According to the method, the calculation and message transmission efficiency of the large-scale multi-label graph is improved.

The invention discloses a topology-enhanced hypergraph learning and multi-graph converter driven drug-related side effect prediction method. The method comprises the following steps: constructing a multi-view heterogeneous graph, a drug-side effect hypergraph and a corresponding line graph; based on a cross-hypergraph and line graph learning strategy, node features of the hypergraph and the line graph are extracted and fused, and a first feature matrix is obtained; integrating node features of the multi-view heterogeneous graph by using an adaptive multi-graph Transform model to obtain a second feature matrix; performing multi-view feature fusion on the first feature matrix and the second feature matrix through a feature-level gating network; and splicing the fusion features and node features of the heterogeneous graph to form drug-side effect node pair feature representation, and performing model training based on a cross entropy loss function to obtain a drug-side effect association model. According to the method, the accuracy of predicting the relationship between the drug and the side effect is improved, and powerful technical support is provided for drug safety evaluation.

The application belongs to the technical field of data processing, and provides a learning ability evaluation model construction method for individual learning of an online learning platform, which comprises the following steps: firstly, collecting behavior data and learning resource metadata of users in a learning process; constructing an individual behavior heterogeneous graph, a knowledge system hypergraph and a knowledge point sequential graph based on the collected data; then, presetting a learning scene attention matrix, combining the knowledge point sequential graph and determining a context perception mechanism; finally, constructing a learning ability evaluation model according to the context perception mechanism, the individual behavior heterogeneous graph and the knowledge system hypergraph. The application realizes accurate modeling of the complex interaction relationship among the user, the learning resource and the knowledge point by constructing the individual behavior heterogeneous graph, the knowledge system hypergraph and the knowledge point sequential graph, deeply fusing the behavior data generated by the user in the learning process with the metadata such as the type, the difficulty and the associated knowledge point of the learning resource, fully showing the systematicness and the dynamics of the knowledge system, and effectively improving the accuracy of the learning ability evaluation.

The application discloses a meter reading scheduling method based on artificial intelligence, comprising: acquiring power grid operation data, constructing an electrical heterogeneous graph and an edge time-varying drive; based on the electrical heterogeneous graph and the edge time-varying drive, using a topological-physical consistent continuous-time distribution prediction model, generating a communication performance prediction distribution containing a confidence band; based on the distribution and the electrical graph, constructing a dynamic communication conflict hypergraph; based on the hypergraph and the prediction distribution, using a hypergraph-constrained graph structure reinforcement learning scheduling model to generate a scheduling plan under the satisfaction of constraints (such as time, concurrency and mutual exclusion). The application improves the accuracy and safety of the prediction model by using physical consistency constraints and conformal calibration, and ensures the robustness and efficiency of the scheduling decision by using the hypergraph-constrained reinforcement learning.

The application provides a hypergraph adaptive sampling method for structure feature reservation, and belongs to the technical field of computer models. The hypergraph adaptive sampling method for structure feature reservation comprises initializing a hypergraph and sampling parameters, the hypergraph comprising a node set and a hyperedge set, setting a sampling ratio, a batch size and a weight adjustment parameter, and calculating a global attribute of the hypergraph; calculating an initial sampling weight of each hyperedge in the hyperedge set based on a hyperedge size, a node degree distribution and a combination ratio; randomly selecting a batch of hyperedges from the hyperedge set according to the sampling weight, adding the batch of hyperedges to a sampling hypergraph, and adding nodes corresponding to the batch of hyperedges to the node set of the sampling hypergraph; calculating a difference value of the sampling hypergraph and the hypergraph in the global attribute, and adjusting the sampling weight according to the difference value. The application can solve the problem that the existing technology cannot realize efficient sampling while maintaining the structure features of the hypergraph.

The invention discloses a firepower attack target distribution method and system based on a multi-constraint dynamic pointer network, and the method comprises the steps: firstly building a hierarchical model which comprises a platform layer, a target layer and a rule layer, carrying out the modeling of a data relation in collected data through employing a hyperedge based on the hierarchical model, and forming a time-varying hypergraph which is updated along with time steps; according to the method, the functions of integrally representing the capacity, the time window, the collaborative grouping, the area limitation and the fair coverage constraint in the same graph space by adopting the time-varying hypergraph and the advancement rule and timely distributing are realized; a primitive-dual online adjustment mechanism is introduced in the aspect of multi-target tradeoff to automatically allocate constraint budget for a Lagrange multiplier according to real-time violation degree, and load variance fairness and regularization are superposed, so that interpretable dynamic balance can be obtained for resource utilization, time window satisfaction, collaborative consistency and coverage balance, a security boundary can be provided, and the real-time performance of the multi-target tradeoff can be improved. And secondary selection of upper-layer strategies or players in the set can be facilitated, and the method is suitable for being widely popularized and used.

This invention discloses a multi-technology fusion prediction method based on hypergraph modeling, relating to the field of computer technology. This method introduces a probability-based null model to correct hyperedge weights, eliminating bias caused by differences in the number of patents across different technology fields, thereby identifying statistically significant multi-technology fusion patterns. Furthermore, the multi-technology fusion prediction problem is transformed into a hyperedge prediction problem within a hypergraph. A random walk algorithm with restart is used to calculate the average similarity among technologies within a technology combination, and this similarity is used as a feature. A logistic regression model is then used to establish a mapping relationship between this feature and the probability of hyperedge formation, thus predicting the probability of technology fusion. This method can effectively identify and predict fusion trends in three or more technology fields, better adapting to the current analytical needs of multi-technology cross-domain integration, and providing data support and decision-making basis for technology innovation layout and industrial policy planning.

The application belongs to the field of graph reconstruction and overlapping community detection, and discloses a hypergraph enhanced semi-supervised graph reconstruction method and an overlapping community detection method, which comprises the following steps: obtaining an attribute graph and a prior matrix, and obtaining an attribute graph adjacency matrix according to the attribute graph; inputting the attribute matrix of the attribute graph and the attribute graph adjacency matrix into a dual attention convolutional encoder to obtain a node representation matrix output by the dual attention convolutional encoder; mapping the attribute matrix of the attribute graph into a hypergraph, performing graph reconstruction on the hypergraph and the node representation matrix, outputting a new attribute graph adjacency matrix, and calculating a graph reconstruction loss function; calculating a semi-supervised loss function based on the prior matrix and the node representation matrix; calculating a modularity maximization loss function based on the attribute graph adjacency matrix and the node representation matrix; and comprehensively training and updating the dual attention convolutional encoder until the training is completed. The application improves the accuracy of graph reconstruction learning and enhances the perception ability of community structure.

The invention discloses a cross-document question and answer method and system based on a sparse hypergraph, and relates to the technical field of knowledge maps, and the method comprises the steps: constructing an initial hypergraph index, taking a document segment as a hyperedge, and connecting all entities in the hyperedge; extracting an entity triple, and distinguishing bridging entities and local entities according to the occurrence frequency of the entities; sparse processing is carried out on the hypergraph, all entities in a 2-hop path related to bridging entities are reserved, redundant local entities are pruned, and a sparse hypergraph knowledge base is generated; 2-hop paths in the knowledge base are converted into natural language problems and vectorized, and a path problem vector library is constructed; after user query is received, trying to directly answer through the matching path question vector library; and if the answer cannot be directly answered, the complex query is decomposed into a minimum single-hop sub-question set, and hyperedges and triples in the sparse hypergraph are iteratively retrieved according to the dependency relationship among the sub-questions to generate answers. The cross-document multi-hop question and answer method effectively improves the accuracy and reasoning efficiency of cross-document multi-hop question and answer.

The invention relates to an efficient hypergraph pattern mining method and device based on an FPGA. In the device, a data preprocessing module is used for organizing and dividing hypergraph data and constructing an active sensing block mass index table, a topology sensing adjacency table and a global hyperedge index table; when the hypergraph data is updated, the dynamic updating module is used for evaluating whether the updating can cause the change of the hyperedge property or the block mass property or not; generating an updating task according to the positioning and evaluation result; updating data structures of the active sensing block mass index table, the topology sensing adjacency table and the global hyperedge index table based on the updating task; the mode compiling module is used for receiving a specified mode and converting the specified mode into an executable instruction sequence, constructing an isomorphic dependency graph and generating an optimal isomorphic matching sequence and an execution plan; the hypergraph mode mining module is used for generating a mode verification task; candidate hyperedges conforming to mode constraints are screened out based on the mode verification task; and embedding the verified candidate hyperedges, and finally constructing a complete embedded instance.

The invention belongs to the technical field of laser surveying and mapping, and particularly relates to a basic surveying and mapping geographic entity rapid extraction method and system based on laser three-dimensional scanning, and the method comprises the steps: S100, carrying out the density-adaptive voxelization downsampling of an original laser point cloud, and calculating geometric, reflection and context features from the original point cloud in a multi-scale neighborhood of each representative point; s200, constructing a hypergraph, generating initial hyperpoints through spectral clustering, iteratively performing semantic classification and geometric consistency check, and dynamically merging or splitting the hyperpoints to obtain a refined hyperpoint set with stable semantic tags; s300, subgraphs are constructed, edge weights reflect geometric and boundary compatibility of combination between the super points, and similar super points are clustered into physical entity instances by calculating a maximum spanning tree and cutting at weak joints; and S400, carrying out geometric fitting and regularization on the point cloud of each entity instance, generating a parameterized model, calculating a spatial relationship between entities, and constructing a global semantic topological graph containing complete attributes and relationships.

The invention discloses an electric power safety regulation cross-text retrieval method and system based on graph retrieval. According to the method, firstly, an electric power regulation document is preprocessed, a knowledge data set is constructed, an entity relation graph and an intra-sentence co-occurrence hypergraph are synchronously constructed according to user problems, and a unified fusion graph matrix is formed through fusion; then, under the constraint of a preset token budget, an information-path collaborative sub-graph retrieval algorithm is adopted, and logically coherent evidence sub-graphs are accurately screened out from the fusion graph; the sub-graph is coded through a graph neural network, multi-dimensional features such as nodes, texts, topologies and types are fused, and a graph-level semantic vector is generated to serve as a structured soft prompt. And finally, splicing the soft prompt and text information, inputting the spliced soft prompt and text information into a parameter-frozen large language model, and driving the model to generate an accurate answer and a clear reasoning path. According to the method, the problems of evidence fragmentation and inference chain incompleteness in multi-hop questions and answers in the field of electric power security are effectively solved, and the accuracy, reliability and interpretability of answers are remarkably improved.

This invention discloses an action recognition method based on pyramid-style segmented dynamic graph construction and ellipsoidal geometric constraint metric learning, comprising the following steps: acquiring a human skeleton action sequence and constructing initial spatiotemporal features; constructing a feature space with ellipsoidal geometric constraints; employing a pyramid-style temporal segmentation strategy, dividing the temporal features into segments of different granularities at different depth levels of the network, constructing local dynamic graphs for each segment, and constructing a global dynamic graph for the entire sequence; adaptively weighting and fusing the generated dynamic hypergraph association matrix, the generated local dynamic graph, and the global dynamic graph with a preset static human topology prior graph to obtain the fused dynamic graph structure; sequentially performing spatial graph convolution and multi-scale temporal convolution on the initial spatiotemporal features to extract deeply fused spatiotemporal features; mapping to the target action category through a classifier to complete action recognition; this invention improves recognition accuracy and generalization ability.

The invention relates to the field of intelligent traffic systems, and discloses an abnormal traffic flow prediction method and system based on multi-scale spatial-temporal feature fusion, and the method comprises the following steps: carrying out the feature enhancement of input traffic data through a multi-dimensional embedded sensing module, and generating a dynamic embedded representation of fusion time, space and traffic state information; and inputting the dynamic embedded representation into a dynamic space-time multi-scale interaction module, and extracting and interacting multi-scale space-time features by enhancing the synergistic effect of a time convolutional network, a graph convolutional network and a hypergraph convolutional network. Through deep integration of time parameters, spatial dependence and traffic state multi-source features by a multi-dimensional embedded sensing module, cooperative operation of an enhanced time convolutional network, a graph convolutional network and a low-rank hypergraph convolutional network in a dynamic space-time multi-scale interaction module is matched; and high-order and non-periodic space-time correlation contained in abnormal traffic flow can be accurately captured.

The invention discloses an anti-information cocoon house recommendation method and system fusing large model semantics and a hypergraph structure. The method comprises the steps that article attribute information and interaction information between a user and an article are obtained through large language model reasoning; constructing an original relation graph; article semantic embedding features and user semantic embedding features are extracted; modeling is carried out on the user behavior sequence through a transform branch, and local fine-grained node representation is captured; constructing a heterogeneous hypergraph structure through hypergraph branches, and capturing high-order node representation of a global topological structure; aligning the local fine-grained node representation and the high-order node representation of the global topological structure through comparative learning; splicing the aligned node representations to obtain a fusion feature; and generating an article personalized recommendation list based on the fused features. According to the method, through LLM-driven semantic enhancement hypergraph construction, double-branch collaborative representation learning and double-vision comparison alignment, the problems of poor semantics, incomplete structure and insufficient diversity of a traditional hypergraph model are fundamentally solved.