26 results about "Sequence graph" patented technology

The invention provides a man-machine marshalling intention recognition method and system based on skeleton trajectory and graph convolution. The method comprises the steps of obtaining real-time personnel behavior video stream data in a man-machine marshalling scene; human skeleton sequence information is extracted from the real-time personnel behavior video stream data, and a personnel skeleton spatial-temporal feature sequence diagram is constructed; and performing joint-limb-behavior three-level feature coding on the personnel skeleton spatial-temporal feature sequence diagram by using a hierarchical graph convolutional network, and outputting a predicted personnel behavior intention.

The invention relates to the technical field of voice visualization, in particular to a visual configuration method and system for a voice social product activity page, and the method comprises the following steps: disassembling voice display content, recognizing page composition and voice attributes, generating a voice fragment set, generating a time sequence chain group according to a trigger type and a sequence, and correspondingly playing a playing instruction and an interaction action. And generating a nested sequence graph and a dependency chain graph, and finally generating a visual configuration scheme. According to the method, structured recognition and unified mapping are completed in voice display content, an ordered trigger chain is constructed in combination with page loading and interaction actions, voice playing and page process synchronization are achieved, interaction coordination is improved through logic correspondence of a playing instruction and an operation action, and time migration and attachment path series playing logic are integrated; the page state change and the voice sequence are fused, redundant content is eliminated, a reusable configuration structure is constructed, the construction efficiency is improved, and the deployment and maintenance cost is reduced.

This invention relates to the field of structural explosion dynamic response calculation, specifically disclosing a heterogeneous graph data representation method for structural explosion dynamic response analysis. The method includes: establishing and validating a high-fidelity finite element baseline model of the structural explosion response, and obtaining a baseline data source for the structural explosion response; establishing and implementing an adaptive sampling algorithm based on physical field gradients to adaptively select key nodes representing the structural dynamic characteristics from the finite element mesh nodes; establishing a multi-attribute graph edge linking and weight quantization method to construct a multi-attribute weighted graph; and establishing an automated construction and storage method for spatiotemporal graph datasets to generate a spatiotemporal sequence graph dataset for training a graph neural network. This invention overcomes the limitations of traditional neural networks on the serialization and meshing of training data, achieving automated, high-fidelity conversion from continuous, heterogeneous finite element simulation data to sparse, discrete graph structure data with well-defined topological relationships, as well as data size simplification and increased physical information density.

The present application relates to the field of graph neural network, deep learning and recommendation system, in particular to a point of interest recommendation method based on time sequence gating graph neural network; the method constructs and trains a user long-term preference model, and uses the trained user long-term preference model to recommend a point of interest for a user; the user long-term preference model comprises a check-in sequence graph construction module, a time sequence gating graph neural network module (TGGNN module), an attention mechanism module and a probability prediction module; the present application represents the check-in activity of a user by constructing a point of interest (POI) check-in sequence graph, realizes the propagation of effective features, designs a time sequence gating graph neural network to dynamically update node vectors by fusing time context information, fully considers the time relationship between nodes in the check-in sequence graph, and can also obtain the complex conversion between different check-in points.

This invention relates to the field of e-commerce recommendation systems, and particularly to an e-commerce recommendation method optimized based on fake review detection. The method includes acquiring user behavior data, user review data, and user-product interaction data; constructing a sentiment-behavior vectorization model, a sentiment-behavior matching detection model, and a sequence graph neural network recommendation model based on user credibility; and obtaining a recommendation list weakened by fake reviews based on the user behavior data, user review data, user-product interaction data, and each model. This invention quantifies the matching degree between user sentiment and behavior and uses it as an indicator of user credibility, increasing the influence of highly credible user behavior on the recommendation results, thereby effectively weakening the influence of fake review data and improving the accuracy of the e-commerce platform recommendation system.

The invention provides a user intention mining method and system based on a double-tower model, a program and a storage medium. The system comprises a data acquisition device, a text graph enhanced representation device, a model training device and an intention mining device. The method comprises the following steps: firstly, mining and converting a user search intention into a dense retrieval problem, and realizing formalized definition of the user search intention and a dialogue text through a Sension-BERT model; then constructing a retrieval architecture based on a double-tower model, performing feature enhancement in combination with a text graph neural network, and capturing text deep association through a semantic graph, a syntactic graph and a sequence graph; meanwhile, an interval constraint negative sample sampling strategy is put forward, effective negative samples are mined from front and back, and the model is trained through a noise comparison estimation method. The method can be deployed at the rear end of each server room, and can widely provide efficient user intention mining support for decision-making systems such as an intelligent medical auxiliary system and the like.

The application discloses a coal rock data generation and permeability prediction method based on deep learning, carries out pretreatment on a coal rock two-dimensional slice image, splices images to obtain three-dimensional binary data; uses a CGAN model to generate image data, uses a Stokes equation and a Darcy law to calculate permeability; parallel feature extraction of CNN and Transform structure is constructed to obtain fusion features for permeability prediction. The application uses a Transform model+CNN model to establish spatial feature selection and sequence feature supplement to predict permeability, ensures the correlation features of sequence graphs, and obtains higher-precision prediction results. Through deep learning, the problem of model underfitting caused by insufficient data is made up, and for the training difficulty of large-size images in the convolution process, a feature based on fusion sampling image slices and non-sampling slices is provided, higher feasibility is brought to model training, model training is facilitated and the demand for training equipment is reduced, and the deployment possibility of the scheme is enhanced.

The invention relates to the technical field of e-commerce supply chains, and discloses an e-commerce supply chain commodity intelligent traceability management method. The method comprises the following steps: accessing a multi-source asynchronous data stream generated in a supply chain circulation complete cycle, and generating a synchronous traceability basic data set after timestamp alignment and cleaning; a space-time trajectory grid of commodity circulation is constructed based on the space-time trajectory grid, and units of the space-time trajectory grid comprise multi-dimensional commodity state vectors. And carrying out state evolution deduction on the grids, generating a commodity circulation behavior sequence map formed by behavior nodes and edges, carrying out multi-modal deep matching on the map and a preset path template, outputting a circulation decision path and decision credibility, and dynamically optimizing packaging protection parameters to generate a protection instruction set according to the circulation decision path and the decision credibility. According to the method, continuous refined perception and intelligent path decision-making of the commodity circulation state are realized, and the accuracy of traceability analysis and the initiative of risk response are improved.

The embodiment of the invention provides a defense strategy determination method and device of an industrial firewall and a computer storage medium. The method comprises the following steps: inputting acquired network intrusion overhead and intruded dynamic feedback information into a Gaussian distribution model to form an intrusion state probability model, forming an intrusion probability graph according to the intrusion state probability model, and constructing an intrusion path sequence graph according to an intrusion path sequence and corresponding intrusion cost, constructing an intrusion feature library by adopting the intrusion probability graph and the intrusion path sequence graph; identifying the intrusion state of the obtained user terminal access behavior data according to the intrusion feature library to obtain intrusion state information; evaluating the security situation of the current industrial network according to the intrusion state information to obtain security situation evaluation information; and determining a defense strategy based on the security situation assessment information. According to the embodiment of the invention, the intrusion state identification accuracy of the behavior access data of the user terminal is improved, the security situation awareness accuracy of an industrial network is improved, and network intrusion is effectively defended.

The invention relates to a software frame rate optimization method and system, computer equipment and a storage medium. The method comprises the following steps: at least synchronously acquiring an operation video and non-visual modal data when target software runs; constructing an interactive response time sequence graph based on the non-visual modal data; analyzing the operation video by adopting a preset double-flow convolutional neural network in combination with the time stream branch and the spatial stream branch to obtain inter-frame motion continuity; adopting a graph neural network, and positioning abnormal resource competition nodes based on the non-visual modal data; and optimizing the operation frame rate of the target software based on the inter-frame motion continuity, the abnormal resource competition node and the interactive response time sequence atlas. The technical problem that an existing frame rate optimization method is poor in software frame rate optimization effect can be solved.

The application provides a human-machine marshalling intention recognition method and system based on skeleton trajectory and graph convolution, and comprises the following steps: acquiring real-time personnel behavior video stream data in a human-machine marshalling scene; extracting human skeleton sequence information from the real-time personnel behavior video stream data, and constructing a personnel skeleton space-time feature sequence graph; performing joint-limb-behavior three-level feature coding on the personnel skeleton space-time feature sequence graph by using a hierarchical graph convolution network, and outputting a predicted personnel behavior intention.

The invention discloses a wireless network dynamic security defense method and device based on risk evolution reasoning. The method comprises the following steps: acquiring log data and network communication data of each node in a target cluster system based on wireless network communication; based on the log data and the network communication data of each node, respectively constructing a time sequence graph structure to obtain a log risk evolution graph and a communication risk evolution graph; fusing the log risk evolution diagram and the communication risk evolution diagram to obtain a risk evolution fusion diagram; performing risk prediction on each node according to the risk evolution fusion graph to obtain a predicted risk value of each node; and according to a preset dynamic risk defense rule and the predicted risk value of each node, determining a target risk defense strategy of each node at different moments. According to the threat identification and prediction based on risk evolution reasoning, the attack evolution trend can be analyzed and predicted in real time, and the corresponding risk defense strategy can be triggered in time, so that more prospective and global security protection can be provided for the system.

The present application relates to the technical field of electronic commerce supply chain, and discloses an intelligent traceability management method for commodity in an electronic commerce supply chain. The method comprises the following steps: accessing multi-source asynchronous data streams generated in the whole cycle of supply chain circulation, generating a synchronous traceability basic data set after time stamp alignment and cleaning; constructing a space-time trajectory grid of commodity circulation based on the synchronous traceability basic data set, wherein the unit of the space-time trajectory grid comprises a multi-dimensional commodity state vector; performing state evolution deduction on the grid to generate a commodity circulation behavior sequence graph composed of behavior nodes and edges; performing multi-modal deep matching between the graph and a preset path template to output a circulation decision path and a decision credibility; and dynamically optimizing packaging protection parameters to generate a protection instruction set. The method realizes continuous and fine perception of the state of commodity circulation and intelligent path decision, and improves the accuracy of traceability analysis and the initiative of risk response.

The application discloses a gait recognition method and system based on multi-modal feature fusion, and comprises the following steps: acquiring a video stream containing a target object; extracting a gait sequence graph of the target object from the video stream; selecting a face optimal frame and a pedestrian optimal frame from the gait sequence graph of the target object; extracting a face feature of the target object from the face optimal frame; extracting a body feature of the target object from the pedestrian optimal frame; extracting a gait feature of the target object from the gait sequence graph; weighting and fusing the face feature, the body feature and the gait feature of the target object to obtain multi-modal fusion features of the target object; and performing identity recognition on the target object according to the multi-modal fusion features of the target object to obtain an identity recognition result of the target object. The accuracy of identity recognition is improved.

This invention discloses a deep knowledge tracking method based on graph neural networks, belonging to the field of smart education. Specifically, the method involves: first, cleaning the real dataset and storing the binary relationship between questions and skills in each interaction data point in a dictionary. Then, for the same student, all their questions are sequenced in ascending time steps, and corresponding exercise sequence graphs and interaction sequence graphs are generated based on the binary relationships. Next, a corresponding graph neural network is constructed to obtain the representation of each graph in the exercise sequence graph and interaction sequence graph. The exercise graph representation at time step t+1 is concatenated with the interaction graph representation at time step t, and the student's learning result at time step t+1 is predicted using a fully connected neural network. Finally, the graph neural network is embedded into an existing Transformer model, and parameters are updated through backpropagation. This invention achieves dynamic graph representation of questions and interaction sequences and utilizes an attention mechanism to update the knowledge state, improving the model's performance and interpretability.

A method and a device for predicting an event to occur in a device are provided. The method includes identifying a plurality of events that occurred in the device; generating a plurality of sequence graphs based on a time-series order of the plurality of events; generating a sequence group graph by combining the plurality of sequence graphs based on relationships between nodes and edges in the plurality of sequence graphs; obtaining a computation graph by sampling the generated sequence group graph based on time-series order information of the plurality of events; and predicting the event to occur in the device using a graph neural network (GNN) based on the obtained computation graph.

The invention provides a multi-layer heterogeneous graph construction method based on a TCP session, and relates to the technical field of network security. The method comprises the following steps: reading original network flow data in a PCAP form and splitting the original network flow data into TCP sessions; sorting the data packets in the TCP session according to the timestamps, and generating data packet nodes; constructing a data packet time sequence edge according to the sequence of the timestamps, and associating each data packet node to construct a data packet sequence diagram; constructing an initial server-side state node and an initial client-side state node; traversing the data packet nodes of the data packet sequence diagram in sequence, and judging whether the states of the server-side state nodes and the client-side state nodes are changed or not; if so, newly establishing a server-side state node and a client-side state node, and constructing a corresponding state transition edge and a corresponding mapping association edge; and if not, constructing a corresponding mapping association edge. According to the method, the key information of the TCP session can be completely described, staged accurate extraction of subsequent features can be conveniently realized, and cross-stage feature dilution can be avoided.

The invention discloses a recommendation method for explicitly dissociating social contact and sequence factors. User decision-making is usually subjected to double influences of social relations and behavior sequences at the same time, an existing recommendation method generally adopts a coupling mode during modeling of the two types of information, complex high-order relations in the information cannot be fully mined, and therefore the accuracy and interpretability of a model are affected. The invention provides a recommendation method (DSSRec) for dissociating social and sequence factors. The method comprises the following steps: firstly, dissociating a social relation and a sequence pattern in a complex interaction graph, respectively constructing a user social graph and an article sequence graph, and designing a multi-channel graph propagation module to capture high-order information on different graphs; and then, constructing a cross-view contrast learning normal form to realize deep dissociation of social influence and sequence influence. And finally, designing a personalized regulation and control mechanism for dynamically regulating the importance of different influence factors during model prediction.

A system log detection method based on a self-attention mechanism graph network first collects log text data and system hardware data from the system's main log file to establish a standard multi-sequence data source. A spatiotemporal graph neural network model is then established, and inter-sequence graph relationships and intra-sequence temporal relationships are built based on the standard multi-sequence data, training optimal parameters. Based on the optimal parameter model, new standard multi-sequence data is used for model inference, and a criterion for judging the inference results is established. Based on the inference results, the region and time of anomaly occurrence are located, and the cause of the anomaly is further analyzed based on hardware data. This invention exhibits good stability and high detection accuracy.

The application discloses a kind of knowledge tracking systems based on heterogeneous graph contrast learning, comprising: data preprocessing module, for collecting educational data and pre-processing;Heterogeneous graph construction module, for constructing educational semantic heterogeneous graph;Graph structure evolution module, for the student behavior stage division of educational semantic heterogeneous graph, constructs time evolution sequence graph;Student behavior subgraph extraction module, for extracting associated structure neighborhood from time evolution sequence graph, generates student behavior subgraph;Hierarchical structure modeling module, for constructing hierarchical semantic heterogeneous graph network, and student behavior subgraph node, edge and subgraph structure hierarchy are modeled by structure representation;Structure contrast optimization module, for designing structure hierarchy contrast learning mechanism and structure disturbance confrontation mechanism, jointly optimize hierarchical semantic heterogeneous graph network;Knowledge tracking module, for realizing multidimensional prediction and evaluation to student knowledge state.The application improves the evaluation capability and generalization robustness of knowledge tracking system.

An encryption and decryption method and apparatus based on uncertain graphs, relating to the fields of computer and network communication technologies, the encryption method includes: A1. Obtaining a plaintext string of length n, constructing a sequence graph and a sequence matrix using the ASCII code value corresponding to each character as a vertex and the difference between the ASCII code values ​​corresponding to adjacent characters as edges; A2. Generating a sequence of uncertain implication subgraphs and an adjacency matrix of uncertain implication subgraphs based on the number of vertices in the sequence graph, and making the non-invertible uncertain implication subgraphs in the sequence graph invertible to obtain a list of adjacency matrices of invertible uncertain implication subgraphs; A3. Dividing the value of the first vertex of the sequence graph by 2... n The quotient and remainder are obtained. The quotient is added to the order matrix to obtain the encryption matrix. The invertible matrix with the same index and remainder is queried from the adjacency matrix list of the invertible uncertain implication subgraph. Finally, the encryption matrix is ​​multiplied by the invertible matrix to obtain the ciphertext. This solves the quantum risk and structural incompatibility problem of traditional encryption and improves the security of the string.

The disclosed embodiments concern methods, apparatus, systems and computer program products for genotyping repeat sequences such as medically significant short tandem repeats (STRs). The methods involve aligning reads to a repeat sequence represented by a sequence graph, and using the aligned reads to genotype the repeat sequence. The sequence graph is a directed graph each including at least one self-loop representing a repeat sub-sequence. In some implementations, the reads are paired end reads, and both mates of each read pair may be used to genotype the repeat sequences. Some implementations can be used to determine degenerate codon repeats. Some implementations can be used to genotype repeat sequences each including two or more repeat sub-sequences. Some implementations can be used to genotype nucleic acid sequences each including at least one repeat sub-sequence and another genetic variant such as an insertion, deletion, or substitution.

The invention discloses a threat detection method and system based on multi-stage attack process matching. The method comprises the following steps: preprocessing heterogeneous traffic in a novel power system, obtaining security alarm information, carrying out association analysis, constructing a coherent attack behavior sequence, and distributing a credibility index for each attack behavior. Uncertainties and conflict evidence present in the alarm data are processed and analyzed. And dynamically generating an attack sequence diagram according to the attack behavior sequence, traversing and reconstructing all possible attack scheme paths according to a time sequence, and selecting a potential attack path with the highest belief value. Through attack stage matching, the most credible attack sequence diagram and the most reasonable attack reduction path are searched, the optimal result pair is identified, and the current killing chain stage of an attacker is determined. According to the method, security analysis is carried out through multi-source information sources, heterogeneous information can be effectively integrated, uncertainty can be processed, an attack scheme can be accurately identified and restored, and complete attack situation awareness is realized.

The invention provides a block chain-based digital base fault evidence storage method and system, the method is realized based on a block chain time sequence fault evidence storage architecture, and the architecture comprises a fault data acquisition layer, a block chain evidence storage layer, an intelligent contract layer and an evidence storage application layer; the method comprises the steps of collecting fault data through a fault data collection layer by means of a digital base hierarchical structure, dividing evidence storage granularity, and generating a standardized evidence storage unit; writing the evidence storage unit into a block chain evidence storage layer, and performing verification by using an intelligent contract layer during and after uplink; and when a user query request is received, calling the evidence storage unit from the verified block chain evidence storage layer through the evidence storage application layer, and generating a fault full-link time sequence graph or a fault evidence storage report. According to the method and the device, full-link tamper-free evidence storage of fault data 'generation-conduction-repair' is realized through a block chain time sequence fault evidence storage framework, and the evidence storage granularity is optimized in combination with a digital base hierarchical structure, so that the fault tracing efficiency is improved.

Embodiments of the present application disclose an attendance recording method and device, electronic equipment and storage medium, the method comprising: obtaining a gait sequence graph of an employee from a collected walking video of the employee; obtaining a gait gradient direction histogram feature and a gait sequence gravity center feature of the employee based on the gait sequence graph; performing feature correlation fusion on the gait gradient direction histogram feature and the gait sequence gravity center feature through kernel canonical correlation analysis, to obtain a gait fusion feature vector of the employee; and recording the attendance of the employee based on an identity label of the employee determined based on the gait fusion feature vector. The technical solution of the present application avoids the loss of effective gait features of the employee when performing feature correlation fusion on the gait gradient direction histogram feature and the gait sequence gravity center feature, solves the problem of incorrect attendance recording caused by the loss of effective gait features of the employee, improves the accuracy of attendance recording, and achieves the purpose of simplifying the attendance process and reducing the workload of the attendance personnel.

The invention discloses a large language model recommendation system and method based on time sequence diagram information enhancement, and belongs to the technical field of artificial intelligence and recommendation systems. According to the method, a two-stage training framework is provided, firstly, in a graph prompt fine tuning stage, parameters of a pre-training language model are frozen, a plurality of sequence interaction graphs are constructed based on user-article interaction historical data, a graph neural network model and a domain alignment module are trained, and graph embedding representation aligned with a semantic space of the language model is generated; and secondly, in a collaborative fine tuning stage, fixing parameters of a graph neural network and a domain alignment module, embedding a graph into an organization, fusing text features to form graph enhancement prompt information, and performing fine tuning on a language model by adopting a parameter efficient fine tuning technology. And finally, forming the recommendation system by the finely-adjusted modules. According to the method, the sequence diagram is embedded into the sequence to be input into the language model, the dynamic evolution of the interest of the user is captured by utilizing a self-attention mechanism, and the accuracy, timeliness, diversity and fairness of recommendation are improved.