159 results about "Multi layer network" patented technology

The invention discloses a voice decoding method, system and device based on electroencephalogram signals and a medium, and relates to the technical field of electroencephalogram signal processing.The method comprises the steps that reading electroencephalogram signals and voice signals in the reading process of a to-be-tested person and imagination electroencephalogram signals in the imagination reading process of the to-be-tested person are collected; inputting the reading electroencephalogram signals and the voice signals into the DRCL to generate electroencephalogram characteristics containing voice information; training the DBM by using the electroencephalogram characteristics as input and using the voice signals as output, and adjusting the DBM by using imaginary electroencephalogram signals to construct a voice synthesizer; a mapping relation between the imaginary electroencephalogram signals and the voice signals is generated through the voice synthesizer, and decoding from the electroencephalogram signals to the voice signals is completed; according to the method, a deep representation correlation learning method is provided, potential correlation between electroencephalogram and voice signals can be deeply mined through a multi-layer network structure, a complex mode which is difficult to recognize by a traditional model is captured, and the voice decoding process is more accurate.

The invention relates to the technical field of node identification, in particular to a network key node identification method fusing improved measurement and propagation influence, which comprises the following steps: setting an edge weight and a cross-layer dependency relationship of an infrastructure network; constructing weighted cross-layer local contact centrality, attenuation cross-layer path centrality and random walk centrality indexes; taking the maximum value of the three adaptive scores as an improved measure of the node; defining metric value standardization; calculating a propagation influence value, standardizing the measurement value, and carrying out weighted fusion on the measurement value and the propagation influence value to obtain a fusion score; carrying out propagation influence value convergence judgment by utilizing an LT linear threshold model; and arranging the nodes based on the fusion score value, and outputting a key node identification result. The method solves the problems that a traditional key node identification method only regards a multi-layer network as a plurality of isolated single-layer networks, lacks measurement of node heterogeneity in the multi-layer network, ignores a coupling relation between network layers and is difficult to be accurately applied to an urban multi-layer infrastructure network.

The invention discloses a spatial domain identification method and device based on multi-modal topological consistency, and belongs to the field of transcriptome spatial domain identification, and the method comprises the steps: constructing a multi-layer network which is in one-to-one correspondence with modal information contained in a biological tissue based on spatial transcriptomics data of the biological tissue; extracting a consensus structure feature shared by the multi-layer network and a specific structure feature specific to each layer of network; constructing a cell consistency network of the biological tissue according to the consensus structural features, the specific structural features and the multilayer network; and performing clustering processing on the cell consistency network to obtain recognition results of different spatial domains in the biological tissue. According to the method, the heterogeneity problem among different modal data can be overcome, and the spatial domain recognition effect is better.

The invention discloses a fine-grained image classification method based on large model enhancement, which guides a model to learn a specific judgment mode of a task by directly introducing prior knowledge. A series of image descriptions are generated by performing question and answer interaction with a multi-modal large language model (MLLM). In order to filter illusion information and redundant content existing in description, a dual-guide text feature optimization module is introduced, and the quality of text features is improved through task-guided feature selection and similarity-guided feature pruning. And finally, adopting a multilayer network structure based on an attention mechanism to realize vision-language fusion for final classification prediction.

The invention belongs to the field of drainage pipeline siltation intelligent diagnosis, and particularly discloses a drainage pipeline siltation intelligent diagnosis method fusing an improved IKAN and a Transform network, and the method comprises the steps: collecting a drainage pipeline inlet and outlet flow and flow velocity data sequence, and carrying out the standardization preprocessing; an improved IKAN module based on adaptive learning is constructed; performing local feature extraction by using a one-dimensional convolutional layer; the feature data is mapped to a high-dimensional token space through a Tokenization module; the method comprises the following steps of: constructing a multi-layer Transform network fused with an improved IKAN (Internet Kalman Area Network); and a double-task learning framework is adopted, and regression prediction results of the deposition thickness and the deposition length and a classification result of whether deposition exists are output at the same time. The optimal performance is obtained when the time step is 15 seconds, the RMSE, the MSE and the MAPE reach 0.282, 0.248 and 0.185 respectively when the noise sample proportion is smaller than 50%, and the method has high robustness and prediction accuracy. According to the method, the accuracy and efficiency of drainage pipeline siltation diagnosis are remarkably improved, and a new technical path is provided for intelligent operation and maintenance of an urban drainage pipe network.

The invention belongs to but is not limited to the technical field of artificial intelligence, and particularly relates to a multi-strategy RAG process optimization method and system for large model questioning and answering, and the method comprises the steps: carrying out the classification processing of a document of a user, and selecting semantic blocks or customizing the sizes of the blocks; receiving a text input by a user; the retrieval is divided into two parts, one part is semantic vector retrieval, and the other part is keyword retrieval; the system further screens and filters candidate text paragraphs in combination with structured meta-information on the basis of vector retrieval; after the preliminary vector retrieval, a bge-ryanker-v2-m3 model is introduced to carry out secondary sorting on the results; the model firstly receives input Prompt and converts the Prompt into vector representation; the vector is processed through a multi-layer Transform network so as to extract and model semantic features in the text; a corresponding output result is generated in combination with specific task requirements defined by Prompt; and outputting an answer generated by the large model for the user question.

The invention discloses a data transmission network optimization method and system for a low-altitude aircraft, and relates to the technical field of unmanned aerial vehicle network optimization, and the method comprises the steps: firstly starting the low-altitude aircraft, setting local communication parameters, and building an initial network registry; periodically broadcasting data by the aircraft, and creating a multi-layer network topological graph; link communication quality prediction is carried out, and relay node path adaptation optimization is carried out according to the link communication quality prediction; a relay node is selected according to an optimization result to establish a real-time data communication path; and finally performing data transmission management based on the selected node and path. The technical problems of poor communication stability, insufficient real-time performance and low data transmission reliability caused by single-dimensional evaluation, insufficient dynamic adaptability and low path optimization efficiency of a traditional communication network optimization method are solved, and multi-dimensional feature fusion, dynamic topology response and multi-target path optimization are achieved. And the technical effects of stability, real-time performance and data transmission reliability of a communication network of the low-altitude aircraft are further improved.

The invention relates to the technical field of ship and ocean engineering, in particular to a global-local dependency cooperative expression ship rolling motion extremely-short-term forecasting method which comprises the steps that an Informer branch and BiGRU-GSA branch parallel architecture is constructed, an encoder of the Informer branch screens key query key pairs through a ProbSparse self-attention mechanism, and the sequence length is compressed through a self-attention distillation mechanism; the decoder adopts generative reasoning and single forward propagation to generate a complete prediction sequence; the BiGRU-GSA branch utilizes a multi-layer BiGRU network to simultaneously capture forward and backward time sequence dependence through a bidirectional gating unit, and extracts time domain fine-grained features; the GSA network constructs local representation of time sequence preference by using two-stage gating linear attention through a gating slot attention model; and the global features of the Informer branch and the local features of the BiGRU-GSA branch are spliced, feature fusion is realized through a full connection layer, a prediction result is generated, and collaborative expression of global-local dependence is realized. According to the invention, high-precision prediction can be carried out on the future motion state of the ship in an extremely short time.

The invention discloses a fuel cell fault diagnosis method based on fusion of a physical model and a neural network, and belongs to the technical field of fuel cell system monitoring and intelligent diagnosis. According to the method, in order to solve the problems that a proton exchange membrane fuel cell stack is complex in operation state, the signal noise of a sensor is large, and a traditional model is difficult to reflect aging and faults in real time, the prior knowledge of a physical model is combined with the time sequence learning ability of a long-short-term memory (LSTM) neural network. The method specifically comprises the following steps: synchronously inputting an actuator or a control signal into a real PEMFC pile and a physical model, and dynamically correcting model parameters by utilizing online parameter identification; performing anomaly detection, filtering and smoothing on a sensor signal, and aggregating with observable and unobservable process variable estimators output by the physical model to form an enhanced feature vector; a multi-scale sliding time window is adopted to construct a multivariable time sequence, the multivariable time sequence is input into a multilayer LSTM network after normalization, and network weights and adjustable parameters of a physical model are updated at the same time through a joint optimization strategy. According to the method, multiple typical faults such as flooding, drying, air depletion and hydrogen depletion can be diagnosed in real time in a classified mode under the dynamic working condition, health indexes such as the performance degradation rate and the remaining life can be output, online intelligent diagnosis and life prediction of the PEMFC pile are achieved, and the method has the advantages of being high in precision, high in robustness, capable of being deployed in an embedded mode and the like.

The invention discloses a social media key user identification system based on information non-uniform propagation characteristics, and relates to the technical field of key user identification, and the system comprises a dynamic module which is used for calling an information propagation model through a weighted directed graph of N time windows, simulating information propagation, obtaining initial influence scores of a user in the N time windows, and obtaining initial influence scores of the user in the N time windows; and a cross-layer module which constructs a multi-layer network, quantifies the influence of the user in different layers in the multi-layer network based on the dynamic influence score, obtains a cross-layer influence score, and performs weighted average on the initial influence score of the user in N time windows to obtain a dynamic influence score. The recognition module is used for calculating comprehensive scores according to the initial influence scores, the dynamic influence scores and the cross-layer influence scores of the users, and the comprehensive scores are arranged in a descending order to recognize key users; through the key user identification method, the social platform can automatically identify key users really having transmission force and guiding force.

The present disclosure relates to the field of artificial intelligence. The present disclosure provides a neural network updating method, a classification method and an electronic device. In the neural network updating method, the neural network comprises a plurality of network layers, at least one of the plurality of network layers comprises a plurality of neurons, and the method comprises: obtaining input data of the neural network; and determining an activation state of at least one neuron in the neural network based on the input data.

The invention discloses a multi-layer network switch based on intelligent routing, and the structure of the multi-layer network switch comprises a slot, a protection device, a heat dissipation net, an indication lamp, a network switch, and a wire plugging port. The adsorption base is stably arranged at the middle position of the lower end in the placement groove under the action of the powerful sucking disc, so that the top of the steering assembly can be connected with the network switch, and meanwhile, the butt joint pressing plate is pressed at the upper end of the network switch in parallel; the attaching piece arranged at the lower end of the butt-joint pressing plate can assist the butt-joint pressing plate to be connected to the network switch in parallel along with the butt-joint pressing plate, the steering assembly is supported by the adsorption base to adjust the angle of the network switch to a corresponding position, and therefore a worker can rapidly and stably insert a corresponding network cable into a corresponding insertion groove. The working intensity of workers is effectively reduced, and the operation convenience of the network switch is improved.

The invention discloses a passive underwater maneuvering target state estimation method based on a long short-term memory network, a program, equipment and a storage medium, and belongs to the technical field of underwater acoustics. The method comprises the following steps: simulating a maneuvering track of a target by adopting a coordinated turning state space model to generate a motion state; performing normalization and time step sampling on the passive acoustic measurement data acquired by the multiple observers, and constructing a time sequence input sequence; inputting the time sequence data into a multi-layer LSTM network to capture a time dependency relationship and nonlinear characteristics of target motion; and training the network through a time back propagation algorithm, and minimizing a mean square error between a state estimation value and a true value. According to the method, verification is carried out in a complex maneuvering scene, and comparison with an interactive multi-model extended Kalman filter and an interactive multi-model unscented Kalman filter is carried out. Results show that the time sequence learning capability of the tracking system can be improved, and the robustness and accuracy of state estimation are enhanced in noise interference and dynamic change environments.

The application belongs to the technical field of natural language processing, and particularly relates to an aspect-level sentiment classification method based on a graph attention network, which comprises the following steps: obtaining word embedding representation of context text in which an aspect word is located; dynamically adjusting the weight of a context word according to the relative position of the context word and the aspect word, and obtaining context semantic features; aggregating syntactic information through an improved graph attention network to obtain syntactic features of the text; using a deep cross network to fuse the syntactic features of the text and the context semantic features to obtain final feature representation; and performing sentiment prediction on the final feature representation through a full connection layer to obtain the sentiment polarity distribution of the aspect word in the text. The application solves the problem of feature information loss that may occur in a multilayer network of the graph attention network, and considers the position information of the context word when extracting semantic features, so that the syntactic features and the context semantic features are fully fused, thereby improving the accuracy of aspect-level sentiment classification.

The application relates to the technical field of public service network, in particular to a public service radius calculation method of adaptive network evolution, which comprises the following steps: dividing the public service into multiple functional units according to the service content provided by the public service, wherein each functional unit comprises at least one service point; constructing a public service network matched with the public service; setting a network evolution mechanism for the public service network to realize adaptive evolution of capacity, connection and modular growth; setting a dynamic adjustment strategy to realize resource elastic scheduling of the public service and optimization of a service coverage range; obtaining the service coverage range, obtaining the service radius of each service point in the service coverage range based on a multilayer network model, combining the number of demand points covered by each service point and resource consumption under each service radius to form an optimal service radius vector set; and determining the optimal coverage radius and the optimal service radius of each service point.

Methods, systems, and apparatuses are provided for managing and routing data in a multi-tier network architecture. The methods include transmitting a data flow comprising a sequence of data packets to a source leaf switch and determining if an intended route exists in a forwarding table stored at the source leaf switch using routing metadata. If the intended route does not exist, a control packet is transmitted to a centralized coordinator switch, which modifies an allocation table to include an updated route. The forwarding table at the source leaf switch is then updated with the updated route, and the data flow is transmitted to its destination.

The invention provides an abnormal transaction monitoring method which can be applied to the technical field of artificial intelligence, the field of information security and the field of financial science and technology. The abnormal transaction monitoring method comprises the following steps: constructing a first multi-layer network based on multi-source transaction data obtained in real time; coupling the first multi-layer network to extract a first aggregated topological adjacency matrix; wherein the first aggregation topology adjacency matrix comprises aggregation transaction characteristics of each layer of network in the first multi-layer network; and monitoring abnormal transactions in the multi-source transaction data through a pre-trained graph network model based on the first aggregation topology adjacency matrix. The invention further provides an abnormal transaction monitoring device and equipment, a storage medium and a program product.

The invention discloses a communication, sensing and computing resource allocation device in a space multilayer network, and the method comprises the following steps: constructing an air-space-ground communication, sensing and computing integrated system, integrating three functions of sensing, communication and computing by an unmanned aerial vehicle of an air-base layer, and performing communication and sensing interference management between the unmanned aerial vehicle and the ground layer through an NOMA-MIMO technology, the tasks can be further unloaded to the low-orbit satellites at the sky level; in order to minimize the weighted energy consumption of the system, a joint precoding, beam forming and resource allocation iterative optimization algorithm is provided; according to the algorithm, an original problem is decomposed into a precoding design sub-problem, a beam forming design sub-problem, an unloading proportion sub-problem and a computing resource allocation sub-problem, and the sub-problems are solved through fractional programming based on quadratic transformation, a weighted minimum mean square error and a continuous convex approximation method. The method shows remarkable superiority in the aspect of optimizing the weighted total energy consumption of the system, and solves the problems of common inductance computing resource allocation, interference management and the like in the spatial multilayer network.

The invention provides a video stream lag intelligent optimization method and device based on AI prediction. The method focuses on the following core links: a data acquisition and feature engineering method covering multilayer network state monitoring, video content deep analysis, user behavior sequence modeling and a real-time data preprocessing algorithm; the AI prediction engine core algorithm comprises an improved Transform architecture, a cross-modal feature fusion mechanism and a multi-scale prediction model; the intelligent optimization decision-making mechanism relates to a reinforcement learning optimization framework, a self-adaptive buffer strategy and an intelligent code rate switching and predictive preloading method; the system architecture and engineering implementation comprises an end-to-end real-time processing architecture, a lightweight model deployment technology, a multi-platform compatible interface and a real-time feedback and tuning mechanism.

The application relates to a network reliability evaluation method and device, computer equipment, a storage medium and a program product. The method comprises the following steps: based on all feasible lower bound vectors of a target network, using a multi-element decision graph to represent all feasible lower bound vectors meeting the service requirement of the target network, and obtaining a feasible lower bound decision graph; performing state space expansion on the feasible lower bound decision graph to obtain a link feasible capacity decision graph; based on a link set of the target network, using a multi-element decision graph to represent the capacity state of each link of the target network, so as to obtain a link state decision graph of each link; merging the link feasible capacity decision graph and the link state decision graph of all links of the target network to obtain a capacity reliability decision graph; and using the capacity reliability decision graph to obtain the capacity reliability of the target network. The method can expand the reliability evaluation method to a multi-layer network, so that the capacity reliability of a complex multi-layer actual network can be accurately described.

The invention discloses a brain cognitive function analysis method based on an EEG-fNIRS cross-modal multilayer network. The method comprises the following steps: firstly, collecting multichannel EEG and fNIRS signals under a working memory training normal form, and preprocessing the EEG and fNIRS signals; source space reconstruction is carried out on the signals, corresponding position information is distributed for EEG and fNIRS source time sequences, and regions of interest are divided based on a Desikan-Killiany atlas; the method comprises the following steps: quantizing decomposable directed mutual information between systems by using IID-PSIT, and constructing an EEG-fNIRS multiplexing network; quantizing cross frequency coupling in the brain function network by using GCSIT, and constructing an EEG-fNIRS full-connection multi-layer network; topological characteristics of a multiplexing network and a multi-layer network are analyzed and extracted by using graph theory parameters, brain function changes under different cognitive loads are revealed, and the accuracy of cognitive load classification tasks is improved.

The invention belongs to the technical field of cell data processing, and relates to a single cell gene characterization system and method based on deep learning. The system comprises a gene information input module, a data processing module, a context sensing interpolation module, a gene feature fusion module, a deep representation learning and joint optimization module and a cell typing and differentiation track inference module. The input module integrates a data source; the processing module preprocesses and enhances the original expression matrix; the interpolation module performs weighted interpolation by using a conditional Gaussian graph model; the gene feature fusion module generates a gene module activity matrix through 1 * 1 convolution and the like; the fusion module fuses the sample meta-information with the active matrix, the sequencing depth and the gene feature vector; the depth representation module captures high-dimensional dependence through a multi-layer Transform network, and generates cell high-order potential representation; the typing module outputs a cell population division and differentiation trajectory based on the deduced pseudo-time trajectory. The method has the beneficial effects that cell expression characteristics are recovered under high sparseness and noise, and a new way is provided for cell type identification and the like.

A connected vehicle includes electronic control units (ECUs) that host onboard cybersecurity modules. The cybersecurity modules are part of a cybersecurity framework that protects the connected vehicle at various logical layers using signature-based modules and rule-based modules that are onboard hosted by the ECUs, and signature-based and anomaly-based modules that are offboard hosted by a computer system that is external to the connected vehicle. Attack techniques of a cyberattack are detected and correlated to validate detection of the cyberattack onboard the connected vehicle and offboard by the external computer system.

The application discloses a respiratory infectious disease population flow network construction and feature mining method and device based on multi-source heterogeneous traffic big data. The method collects multi-source traffic travel data such as railways, civil aviation, highways and mobile signaling and pre-processes, constructs a time-space correlation database, and generates multiple standard data tables through cleaning, desensitization, time-space alignment standardization processing. Based on the standardized data, the network topology parameters are defined, the multi-traffic mode OD flow matrix is generated, and the multiple verification and calibration are completed, the civil aviation, railway and highway hierarchical networks are respectively built, the network edge weight is fused through node alignment and interlayer coupling, the super-adjacency matrix is constructed, the multi-dimensional topological features of the multi-layer network are mined, and finally the exclusive feature data set suitable for respiratory infectious disease cross-regional transmission prediction is generated. The scheme can efficiently fuse standardized multi-source heterogeneous data, accurately construct the population flow network, and effectively improve the accuracy and timeliness of the infectious disease transmission risk assessment.

An automatic detection method for CBCT cephalometric landmarks based on multi-geometric guidance and specific perceptual coding includes the following steps: Step S1, downsampling the 3D CBCT image and obtaining a preliminary coordinate set through a coarse localization network; Step S2, cropping image blocks centered on the coordinates and extracting local features through a visual encoder containing a shared basic encoder and a low-rank adapter; Step S3, calculating the relative position matrix of the landmarks, encoding spatial relationships using radial basis functions, and constructing a multi-anatomical heterogeneous map; Step S4, inputting visual and edge features into a multi-geometric guidance Transformer, fusing global constraints and updating features using an attention mechanism; Step S5, extracting directional geometric relationships using spherical harmonic functions to construct higher-order update terms, and dynamically updating the heterogeneous map using a gated residual mechanism; Step S6, predicting coordinate offsets through a multi-layer network and performing iterative optimization to output high-precision 3D coordinates. This method significantly improves detection accuracy and robustness.

Techniques are described for optimizing multipaths of a segment routing-enabled network. For example, a computing device is configured to: for each link in a network layer of a multi-layer network, compute a usage (metric) of the link by all paths of a first plurality of multipaths provisioned in the network layer to compute a total usage by the first plurality of multipaths, the first plurality of multipaths having been computed and placed to a model of the network layer in a first order; compute a second plurality of multipaths, wherein the second plurality of multipaths are computed and placed, to the model of the network layer, in a second, different order; and in response to determining that the total usage by the second plurality of multipaths is less than the total usage by the first plurality of multipaths, provision the second plurality of multipaths in the network layer.

Systems and methods for dynamic bandwidth allocation and jitter handling in multi-tiered satellite networks such as virtual communication networks and / or physical communication networks. The system includes bandwidth distributing (BD) unit, in-route group managers (IGMs), in-route bandwidth manager (IBM). The system functions in multi-tier network entity management mode, and multi-beam management mode. The system operates by periodically collecting bandwidth reports from IGMs. The BD unit analyzes data in report and compares aggregated bandwidth demand with pre-defined thresholds. If thresholds are exceeded, system switches to multi-tier network entity management mode. In this mode, BD unit determines individual bandwidth adjustments for virtual network and beams and allocates bandwidth for individual devices within virtual network. Further, BD unit transmits this information to IBM, which applies allocations, and receives and transmits a scaling factor back to IGM indicating utilization of allocated bandwidth. Further, BD unit switches back to multi-beam management mode based on comparison result.

The invention discloses a methanol-to-olefin production safety anomaly detection method based on an LSTM algorithm. The method comprises the following steps: acquiring methanol-to-olefin real-time process parameters and control instruction data; preprocessing the process parameters and the control instruction data to obtain a training set and a test set; constructing a reconstruction model based on the multilayer LSTM network, and carrying out training learning on the reconstruction model by adopting the training set to obtain a joint normal model; inputting the test set into the joint normal model, and calculating and outputting a reconstruction error of the test set and the test set in real time; and establishing a multi-stage abnormal response mechanism. According to the method, by analyzing the joint change mode of the multi-dimensional time sequence data, the complex abnormal condition can be accurately identified, the high-efficiency detection of the potential safety hazard in the olefin production process is realized, and the defect that the traditional method is difficult to fuse the multi-dimensional information and accurately identify the complex abnormal condition is overcome.

The invention discloses a point cloud registration method and device based on multi-modal space perception and electronic equipment. The method comprises the following steps: acquiring an RGB-D image of a target scene and a pair of three-dimensional point clouds; extracting image features from the RGB-D image by using an image feature extractor; encoding the three-dimensional point cloud into three-dimensional point cloud features by using a 3D feature encoder; projecting the three-dimensional point cloud to a plurality of planes to obtain a plurality of plane features; fusing the three-dimensional point cloud features, the plurality of plane features and the image features by using a multi-layer network based on an index relationship between the features to obtain initial multi-modal fusion features; calculating a point confidence coefficient and a plane confidence coefficient based on the three-dimensional point cloud and the plurality of plane features; on the basis of the point confidence coefficient and the plane confidence coefficient, pruning is carried out on the initial multi-modal fusion features, and a point corresponding relation used for registration is obtained; and point cloud registration is carried out based on the point corresponding relation, so that the registration precision is improved.

This invention discloses a forward-backward iteration power flow optimization calculation method, medium, and equipment for active distribution networks, relating to the fields of distribution network analysis, operation, and control technology. It constructs an active distribution network based on the physical connection structure of the network and hierarchically divides it into multi-layer networks. The number of iterations is predicted based on the parameters and corresponding correlation coefficients of the multi-layer network to obtain the minimum number of iterations. The PV node sensitivity impedance matrix is ​​calculated based on the attributes of all nodes in the network. After performing forward-backward iteration on the active distribution network, it is determined whether the current iteration count has reached the minimum number of iterations. If it has, the iteration stops and the calculation result is obtained based on node voltage convergence or the upper limit of the iteration count; otherwise, the iteration continues. By predicting the minimum number of iterations, employing an efficient PV node sensitivity impedance matrix construction method, and dynamically optimized relaxation factors, the efficiency and robustness of forward-backward iteration power flow calculation for active distribution networks are significantly improved while ensuring computational accuracy.