Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

160 results about "Singular value decomposition method" patented technology

Singular value decomposition (Singular Value Decomposition, SVD) is the decomposition of a real matrix in order to bring it to a canonical form. Singular decomposition is a convenient method when working with matrices.

Three-dimensional electromagnetic spectrum map construction method based on tensor completion

The invention relates to the technical field of data visualization, and particularly discloses a three-dimensional electromagnetic spectrum map construction method based on tensor completion, which comprises the following steps: S1, acquiring a to-be-detected area in a simulation scene, and acquiring spectrum data in the to-be-detected area to obtain a missing data set; s2, based on a tensor singular value decomposition method and the missing data set, obtaining a complemented tensor; and S3, constructing a three-dimensional electromagnetic spectrum map based on a Poisson surface reconstruction algorithm and the complemented tensor.
Owner:PLA PEOPLES LIBERATION ARMY OF CHINA STRATEGIC SUPPORT FORCE AEROSPACE ENG UNIV

Path planning method of laser welding head, medium and system

The invention provides a path planning method of a laser welding head, a medium and a system, and belongs to the technical field of laser welding, the laser welding path planning method provided by the invention comprises the following steps: firstly collecting image information of a to-be-welded part, predicting a welding effect through a deep learning model, and establishing a scientific welding gain matrix evaluation system; a singular value decomposition method is adopted to intelligently segment a welding seam, welding feature vectors are automatically extracted, and a process parameter space is established. And the system constructs a multi-objective optimization function based on the welding quality comprehensive score, and generates an optimized welding path. The welding process image and the quality thermodynamic diagram are collected in real time, the feature similarity with the preset effect is calculated, dynamic optimization of the path is achieved, the intelligent welding path is finally output, the limitation of traditional artificial experience path planning is broken through, and the path planning efficiency is improved. The technical problem that in the prior art, laser welding often depends on manual experience to preset the welding path, and the welding path is not accurate enough is solved.
Owner:BEIJING NANCAL RUIYUAN DIGITAL TECH CO LTD

Structural deformation non-contact dynamic measurement method and system based on ultra-sensitive optical flow method, processing equipment and storage medium

The invention provides a structural deformation non-contact dynamic measurement method and system based on an ultra-sensitive optical flow method, processing equipment and a storage medium. The method comprises the following steps: acquiring an image of a to-be-measured structure; calibrating a scale parameter p according to the physical size of the structure and the number of pixels occupied by the structure image; constructing an effective pixel original gray scale space-time matrix by using pixels with relatively large image gray scale gradient generated by a structure edge (or surface natural texture); a structure deformation mode is calculated and recognized through a singular value decomposition method, and then a weight matrix based on a deformation shape is constructed; performing weighted average filtering on an original gray scale space-time matrix by using the weight matrix, and restoring a gray scale value decimal part lost due to digitization by using a noise signal naturally generated in an imaging process and a signal jitter principle; and finally, calculating to obtain a pixel displacement time history of each effective pixel point by using a gradient-based optical flow method, and converting through a scale parameter p to obtain a physical displacement time history.
Owner:杨永超 +1

Data processing method of seabed-based monitoring equipment

The invention provides a data processing method for seabed-based monitoring equipment, which belongs to the technical field of electric digital data processing, and comprises the following steps: normalizing acquired data, removing abnormal values, extracting key features by using a singular value decomposition method, and segmenting by using a self-adaptive sliding time domain window to form time sequence data blocks. According to the method, a CUDA parallel processing flow is creatively used to construct a disorder matrix and a drift matrix, and a disorder mode feature vector, a drift principal component and an abnormal feature matrix are fused through a tensor decomposition technology to construct a seabed environment state tensor model. Meanwhile, a drift rule in a long-time sequence is learned by utilizing a deep sea environment perception attention network model, real-time monitoring data is corrected and compensated, and finally a seabed state evaluation standard is established by adopting a hierarchical clustering algorithm, so that the key technical problem that environment change and equipment drift in seabed monitoring data are difficult to distinguish is effectively solved.
Owner:青岛道万科技有限公司

Water vapor chromatography modeling method and device based on GNSS-MET

The invention discloses a water vapor chromatography modeling method and device based on GNSS-MET, and belongs to the technical field of atmospheric water vapor information inversion. According to the method, satellite signals and meteorological parameters are obtained through a GNSS-MET receiver, and the oblique path water vapor content (SWV) is calculated to serve as chromatography input; densely and uniformly distributed observation stations are selected to construct a chromatography area, and voxel grids are divided; a horizontal constraint is constructed by using Gaussian distance weighting, a vertical constraint is constructed in combination with a water vapor vertical index distribution characteristic, and ERA5 reanalysis data is introduced as a prior constraint, so that the limitation of a traditional sounding data constraint is solved; and resolving the tomographic equation through a singular value decomposition (SVD) method to obtain three-dimensional water vapor density distribution. According to the method, the high temporal-spatial resolution characteristic of ERA5 data is utilized, the precision and reliability of water vapor chromatography are remarkably improved, and the method is suitable for the fields of extreme weather prediction, meteorological monitoring and the like.
Owner:AEROSPACE INFORMATION RES INST CAS

Excavation-while-excavation seismic signal feature extraction method based on local frequency

The invention discloses a method for extracting while-digging seismic signal features based on local frequency, and the method comprises the steps: obtaining while-digging seismic data, and carrying out the preprocessing of the obtained while-digging seismic data; performing time-frequency analysis according to the preprocessed data to generate a time-frequency spectrum, and constructing a local frequency weight matrix according to the time-frequency spectrum; performing weighted singular value decomposition on the seismic data while digging according to the local frequency weight matrix to obtain an optimized decomposition matrix; and performing low-rank reconstruction on the optimized decomposition matrix, and extracting the characteristics of the seismic signal while digging. On the basis of a local weighted singular value decomposition method, local frequency energy weights of reflected signals are extracted through time-frequency analysis, and singular value decomposition is guided to strengthen effective components and suppress noise in low-rank approximation. According to the method, self-adaptive enhancement and noise separation of signals are achieved, the signal-to-noise ratio and fidelity of coal-rock interface reflection characteristics are remarkably improved while the calculation efficiency is guaranteed, and reliable technical support is provided for excavation-following detection under complex geological conditions.
Owner:INNER MONGOLIA RESEARCH INSTITUTE CHINA UNIVERSITY OF MINING AND TECHNOLOGY (BEIJING)

Regional drought and flood distribution prediction method and system based on big data and multiple models

The invention discloses a regional drought and flood distribution prediction method and system based on big data and multiple models. Historical rainfall and circulation sea temperature field data of a research area are collected, historical samples of typical drought and flood years are screened, SPI indexes of the typical drought and flood years are calculated, and time and space characteristics of the typical drought and flood years are obtained; constructing a typical drought and flood annual conceptual physical model and a regression clustering forecasting physical model forecasting map, and obtaining a climate background abnormal characteristic value and a drought and flood grade forecasting value of the research area; constructing a precipitation field optimization prediction model based on a singular value decomposition method; and constructing a drought and flood distribution prediction model of the research area based on the improved RegCM4 model to obtain drought and flood distribution of the research area. Compared with a traditional drought and flood prediction method, the method can provide a technical basis for drought and flood long-term rainfall prediction in the future, and improves the long-term rainfall prediction level.
Owner:CHINA YANGTZE POWER

Image restoration method based on adaptive weighted tensor completion

The invention provides an image restoration method based on adaptive weighted tensor completion, and relates to the technical field of image processing and application, and the method comprises the steps: obtaining to-be-restored image data, and carrying out the tensor of the to-be-restored image data, and obtaining input tensor data; constructing a tensor completion model based on an adaptive weighted tensor nuclear norm; wherein a weight matrix in the tensor completion model can be adaptively updated along with input tensor data; and based on an alternating direction multiplier method or an approximate singular value decomposition method based on tensor QR decomposition, solving the tensor completion model, and outputting restored tensor data to realize image restoration. According to the scheme, the image restoration quality can be improved.
Owner:NINGXIA UNIVERSITY

Hybrid expert model training and merging method based on efficient fine tuning of parameters

The invention discloses a hybrid expert model training and merging method based on efficient fine tuning of parameters. The method comprises the following specific steps: S1, firstly, training a base model; s2, performing extraction and initialization LoRA on a matrix principal component of a main network layer by using a singular value decomposition method, and leaving a residual error in an original matrix; s3, for a plurality of fields, training a plurality of LoRA branches; and S4, after training is completed, adding a gate control layer, taking a plurality of LoRAs as activation experts of the MoE, taking the residual matrix as a side branch, and training the gate control layer by using a small part of merged data. According to the method, the computing power cluster scale is remarkably reduced, the convergence efficiency can be remarkably improved through the singular value decomposition method, the final parameter scale is remarkably reduced due to the fact that the final MoE uses the LoRA structure, and the model effect is better.
Owner:SHENZHEN YUHANG TECHNOLOGY CO LTD

Anchoring fracture rock deformation displacement lightweight prediction method and equipment

The invention discloses an anchoring crack rock deformation displacement lightweight prediction method and equipment, and belongs to the technical field of intelligent early warning of geotechnical engineering disasters. Constructing a prediction model by adopting a random configuration network method; partial hidden nodes with low contribution degree in the prediction model are eliminated by using a singular value decomposition method, the number of the hidden nodes in the prediction model is reduced, and the prediction model is lightened; excellent new hidden nodes are searched through a random hidden node increasing and replacing technology to replace original hidden nodes in the prediction model, a lightweight prediction model is reconstructed, and the prediction precision is improved; the lightweight prediction model is utilized to predict the deformation displacement of the anchoring crack rock; the method has simple steps, and can accurately predict the deformation displacement of the anchoring crack rock based on the position of the anchor rod and the load level of the anchoring crack rock, thereby ensuring the safety of geotechnical engineering.
Owner:CHINA UNIV OF MINING & TECH

Fine tuning of diffusion-based generative neural networks for text-to-image generation using singular value decomposition

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for trimming diffusion-based generative neural networks in a compact parametric space for text-to-image generation. In one aspect, a method performed by one or more computers for trimming a diffusion-based generative neural network to obtain a trimmed version of the diffusion-based generative neural network is described. The method comprises: for each of a plurality of neural network layers of a diffusion-based generative neural network: obtaining an initial weight matrix comprising a plurality of pre-trained weights parameterizing the neural network layer; performing singular value decomposition on the initial weight matrix; and reparameterizing the neural network layer with a new weight dependent on the spectral shift; and training a spectral shift of each of the plurality of neural network layers of the diffusion-based generative neural network to obtain a trimmed version of the diffusion-based generative neural network.
Owner:GOOGLE LLC

Mobile user track prediction method and system

The invention provides a mobile user trajectory prediction method and system, and relates to the technical field of computer information communication and service computing, and the method comprises the steps: obtaining historical spatial-temporal trajectory data of a mobile user, introducing social relation data, and constructing a heterogeneous graph; performing data enhancement on the heterogeneous image from a structure layer and a feature layer through a feature extraction model based on comparative learning, capturing a dependency relationship and interest preference in spatial-temporal trajectory data, and extracting trajectory features; on the basis of the track features, interest point prediction is carried out; the structural layer is used for extracting structural information in the heterogeneous graph by adopting a singular value decomposition method and reconstructing the heterogeneous graph; the feature layer is characterized in that noise is added to a feature space in a process of carrying out comparative learning on an original heterogeneous image and a reconstructed heterogeneous image through a lightweight image convolutional neural network; according to the method, by introducing social relation data and a comparative learning mechanism for performing data enhancement from two levels, trajectory features are accurately extracted, and the problem of spatio-temporal data sparsity is relieved.
Owner:SHANDONG UNIV

Wide-area distributed photovoltaic power generation efficiency evaluation method

The invention provides a wide-area distributed photovoltaic power generation efficiency evaluation method, and relates to the technical field of photovoltaic power generation efficiency evaluation, and the method comprises the steps: obtaining the reanalysis time series data of the solar radiation of an area where a distributed photovoltaic power station is located, calculating a feature vector through employing an empirical orthogonal function decomposition method, and calculating the power generation efficiency of the distributed photovoltaic power station; a meteorological resource representative point set of final site selection is screened out in combination with site selection conditions; collecting the earth surface solar total radiation measurement data of the meteorological resource representative point set, and correcting the reanalysis time sequence data through a singular value decomposition method; matching a solar radiation reanalysis grid for the distributed photovoltaic power station, and evaluating the distributed photovoltaic power generation efficiency through a physical method and a statistical method based on the reanalysis time sequence data after correction processing to obtain an evaluation value. The method has wide applicability, and resource distribution analysis can be realized only by using a public weather reanalysis data set covering the whole world.
Owner:HUANENG JIANGSU COMPREHENSIVE ENERGY SERVICE CO LTD

Method, system, device and medium for measuring temperature and reducing noise of distributed optical fiber

To provide a method, a system, a device, and a medium for measuring temperature and for reducing noise of a distributed optical fiber, which relate to the technical field of noise reduction of a distributed sensing system.SOLUTION: A method includes the steps of: obtaining an anti-Stokes signal and a Stokes signal of a photovoltaic assembly collected by temperature measurement of a distributed optical fiber and a noise reduction system, calculating a temperature ratio signal; dividing the temperature ratio signal into a plurality of temperature ratio signal sub-blocks; performing initial iterative noise reduction processing on original sub-matrices corresponding to different temperature ratio signal sub-blocks for different iteration times by using a singular value decomposition method combining fusion functions, to obtain noise-reduced temperature data sub-blocks corresponding to each temperature ratio signal sub-block; determining an iteration time with a good noise reduction processing effect and a reference sub-block, and performing iterative noise reduction processing on a non-reference sub-block based on the iteration time with the good noise reduction processing effect, splicing the non-reference sub-block into noise-reduced temperature fusion data, and calculating a final temperature demodulation result.SELECTED DRAWING: Figure 1
Owner:CHINA THREE GORGES RENEWABLES (GRP) CO LTD

Roadway roof stability evaluation method based on while-drilling parameter response characteristics

The invention relates to a roadway roof stability evaluation method based on while-drilling parameter response characteristics. The method comprises the steps that any roof is selected in a roadway, and while-drilling parameters of a drilling machine in the roadway roof drilling process are continuously collected through a sensor; noise reduction is conducted on the while-drilling parameters through a singular value decomposition method, a while-drilling data set after noise reduction is obtained, unit volume rock drilling energy consumption is analyzed, and the rock uniaxial compressive strength is obtained based on unit volume rock drilling energy consumption inversion; according to the drilling energy consumption or drilling images of different drilling depths, the thickness of a complete rock stratum of the roadway roof, the number of fractures and the width of the fractures are analyzed; according to the thickness of the complete rock stratum, the accumulated complete rock stratum proportion of the roadway roof is analyzed, and according to the rock uniaxial compressive strength of the roadway roof, the thickness of the complete rock stratum, the number of fractures, the width of the fractures and the accumulated complete rock stratum proportion, the stability evaluation result of the roadway roof is determined. Therefore, effective grading evaluation can be performed on the top plate stability simply and quickly.
Owner:CHINA UNIV OF MINING & TECH

Cable segmentation wave velocity acquisition method, device and system based on Prony algorithm, and medium

The invention provides a Prony algorithm-based cable segment wave velocity acquisition method, device and system, and a medium, and the method comprises the steps: testing a to-be-tested cable, and obtaining a cable signal transfer function; on the basis of the cable signal transfer function, in combination with cable joint distribution, constructing a cable signal approximation function based on a Prony method; constructing a linear prediction model of the cable signal based on the cable signal approximation function; performing denoising processing on the linear prediction model based on a singular value decomposition method to obtain a denoised linear prediction model; solving the denoised linear prediction model to obtain an attenuation coefficient; and calculating the segmented wave velocity of the cable based on the attenuation coefficient. According to the method, the Prony estimation method is combined with the singular value decomposition noise reduction algorithm, high-precision extraction of attenuation constants and segmented wave velocity decoupling are achieved, and then the electrical distance positioning precision is improved.
Owner:STATE GRID HUBEI ELECTRIC POWER RES INST

Truss type load-bearing structure time domain dynamics topological optimization method and system based on intrinsic orthogonal decomposition method

PendingCN121351475AGeometric CADDesign optimisation/simulationTime domainGeometric control
The invention discloses a truss type load-bearing structure time domain dynamics topological optimization method and system based on an intrinsic orthogonal decomposition method, and relates to the technical field of truss type load-bearing structure topological optimization. The method is provided for solving the problems of high calculation power and the like caused by large-scale space-time response iteration of original transient problems and intensive calculation of dual problem sensitivity analysis in the prior art. According to the method, a local geometric control strategy for the cross interference problem of rod piece units is introduced, and a transient dynamics full-order topological optimization model of the truss structure is established through an equal geometric stiffness diffusion method. Based on an intrinsic orthogonal decomposition method, a transient dynamics reduced-order model of a truss structure is constructed, a reduced-order model base is updated in real time through an incremental singular value decomposition method, and then efficient solving of a transient dynamics equation and an adjoint equation in sensitivity analysis is achieved. And updating an iterative design variable by adopting a penalty function method, and seeking the optimal layout design of the truss structure. The invention provides a time-domain dynamics topological optimization method of the truss type load-bearing structure based on an intrinsic orthogonal decomposition method, and the method can be used for efficiently solving the dynamics layout optimization problem of the truss structure under the action of a transient load.
Owner:HARBIN UNIV OF SCI & TECH

Canopy morphology 3D feature analysis method based on lightweight unmanned aerial vehicle and point cloud singular value decomposition

The invention discloses a canopy morphology 3D feature analysis method based on a lightweight unmanned aerial vehicle and point cloud singular value decomposition, and belongs to the field of plant morphology analysis. The method comprises the following steps: carrying out high-precision three-dimensional reconstruction on a soybean canopy structure by adopting an unmanned aerial vehicle to carry an RGB lens and combining with an SfM-MVS algorithm, introducing a singular value decomposition (SVD) method for a point cloud coordinate matrix based on a label method, and establishing nine canopy morphological 3D features under multiple scales as new explainable features of a crop plot level. Results show that the global canopy 3D features have significant correlation (plt; plt) with the yield, lodging, canopy height and vegetation index of soybeans; = 0.05). More detailed structure information of the canopy surface can be extracted based on 3D feature extraction of a local range, and characterization of differences of more growth periods of soybeans is facilitated.
Owner:CHINA AGRI UNIV +1

Rural environment assessment method and system based on modern human settlement environment conditions

The invention discloses a rural environment assessment method and system based on modern human settlement environment conditions, and the method comprises the steps: collecting, obtaining and processing rural environment basic data, and obtaining a target data set; performing rural environment feature zoning based on the target data set, a sampling space clustering algorithm and a spatial analysis technology, extracting environment feature indexes of environment feature regions, and performing spatio-temporal dynamic evaluation on the environment quality condition of each environment feature region by adopting a singular value decomposition method to obtain an environment quality evaluation result; and establishing a rural environment evaluation early warning mechanism, performing threshold judgment on an environment quality evaluation result, performing early warning in time and taking corresponding control measures when the environment quality index of a certain region exceeds a preset threshold, and dynamically updating a rural environment characteristic zoning result and an environment quality evaluation result according to an environment change condition. Scientific division, comprehensive evaluation and dynamic early warning of the rural environment are realized, and data support and scientific basis are provided for rural environment management and decision making.
Owner:INSTITUTE OF ENVIRONMENT AND SUSTAINABLE DEVELOPMENT IN AGRICULTURE CAAS

Identification method and system for static voltage analysis parameter model of electrolytic aluminum load

Disclosed in the present invention are an identification method and system for a static voltage analysis parameter model of an electrolytic aluminum load. The method comprises: acquiring historical parameter data generated when an electrolytic aluminum load is connected to a power grid under the normal working condition, and selecting parameters to be identified; on the basis of said parameters, establishing a static voltage analysis model of the electrolytic aluminum load; on the basis of the static voltage analysis model of the electrolytic aluminum load in combination with a singular value decomposition method and a preset constraint condition, optimizing the static voltage analysis model of the electrolytic aluminum load, and acquiring a static voltage analysis model of the electrolytic aluminum load subjected to parameter identification; and using a BP neural network method to verify the static voltage analysis model of the electrolytic aluminum load subjected to parameter identification. A verification objective function is established on the basis of the difference value between data of two models and the actual measurement data, and a parameter model identification result of the electrolytic aluminum load is verified on the basis of the value of the verification objective function, so that the static voltage analysis model of the electrolytic aluminum load is accurately and efficiently established.
Owner:YUNNAN POWER GRID CO LTD +1

A 6D pose grasping method for robotic arm based on neural network

The invention discloses a 6D posture grasping method of a robot arm based on a neural network, comprising: respectively acquiring RGB and depth video sequences through an industrial depth camera; building an improved YOLOV5 algorithm network, the network comprising a backbone network, a network layer and a prediction layer; training the algorithm network and a preset PVN3D-Tiny network based on a preset loss function; sending the RGB video sequence to the trained algorithm network to obtain an output feature map; based on an ROI area of ​​a target to be grasped in the output feature map and a depth video sequence, using a preset trained PVN3D-Tiny network to obtain predicted key points on the surface of a target to be grasped, and using a singular value decomposition method to calculate 6D posture information of a target to be identified; acquiring obstacle information, based on the 6D posture information of the target to be identified, combining a seventh-order polynomial trajectory planning algorithm and a genetic algorithm to obtain a planning trajectory, combining the obstacle information to perform obstacle avoidance on the robot arm to obtain an optimal trajectory, thereby improving the accuracy and speed of target detection and grasping.
Owner:HUNAN INSTITUTE OF ENGINEERING

Electric vehicle charging station recommendation method and system

An electric vehicle charging station recommendation method and system, applied to the related field of charging station recommendation. The method comprises: acquiring user charging features and influence degree values of user charging evaluation features on user charging preferences, wherein the user charging evaluation features are constructed on the basis of the user's evaluation information on charging stations (S1); calculating the user's comprehensive scores for different charging stations on the basis of the influence degree values, and constructing a user preference matrix by using the user's comprehensive scores for different charging stations as elements (S2); and decomposing the user preference matrix by means of a singular value decomposition method, acquiring potential matching relationships between the user and charging station features, and recommending a charging station to the user on the basis of the potential matching relationships (S3). The method solves the problem in the prior art that charging stations cannot be personalizedly recommended for different users, thereby improving user experience.
Owner:STATE GRID ELECTRIC VEHICLE SERVICE CO LTD +1

Mass airborne LiDAR point cloud rapid registration algorithm fusing color features

The invention discloses a mass airborne LiDAR point cloud rapid registration algorithm fused with color features, in the registration process, a point cloud rasterization strategy is used, color information of point clouds is fused into corresponding feature point extraction, a SuperPoint and SuperGlue combination mode is selected, feature point extraction and matching of large-difference point cloud data are realized, and the registration accuracy is improved. A singular value decomposition method is designed to be combined with a random sampling consensus algorithm, rotation translation matrix calculation is completed, and a comprehensive evaluation index system is established. The method can be effectively used for high-precision registration of outdoor large-area massive airborne LiDAR point cloud data, in addition, the algorithm has good tolerance to the number of point clouds, by combining downsampling, even if the number of point clouds reaches more than ten million levels, the registration time can still be kept about 1 second, and the algorithm has good adaptability and can be used for high-precision registration of the large-area massive airborne LiDAR point cloud data. The method can also be applied to registration between point cloud data generated by point cloud without color information or images with large differences.
Owner:Chinese People's Liberation Army Cyberspace Force Information Engineering University

Trolley positioning method, device and equipment based on laser radar and storage medium

The invention provides a trolley positioning method and device based on a laser radar, equipment and a storage medium, and relates to the technical field of tunnel engineering. The method comprises the following steps: acquiring an attitude transformation matrix between a laser radar and a trolley, and a target center first coordinate of each target; obtaining point cloud data of the laser radar, and obtaining a target center second coordinate of each target according to the point cloud data; according to the target center first coordinate and the target center second coordinate of each target, a first conversion matrix between the radar coordinate system and the tunnel coordinate system is obtained through a singular value decomposition method, and according to the first conversion matrix and the attitude conversion matrix, a second conversion matrix between a trolley coordinate system where the trolley is located and the tunnel coordinate system is obtained; and positioning the trolley according to the second conversion matrix. According to the method, the positioning precision of the trolley in the tunnel is improved on the premise of high universality.
Owner:CHINA RAILWAY CONSTR HEAVY IND

Heterogeneous air quality data fusion method based on sparse matrix decomposition

The invention discloses a heterogeneous air quality data fusion method based on sparse matrix decomposition. The method comprises the following steps: S1, collecting air quality data; s2, preprocessing the collected air quality data, and constructing a preliminary fusion matrix; s3, marking missing data in the preliminary fusion matrix, and constructing a sparse matrix; s4, carrying out matrix decomposition by adopting a mode of combining an improved singular value decomposition method and a matrix completion method based on low-rank representation, filling missing data, and carrying out feature weighted optimization on a matrix decomposition result by utilizing an attention mechanism; and S5, error measurement and consistency check are carried out. According to the method, sparse matrix decomposition and low-rank completion technologies are adopted, the fusion process of the air quality data is optimized in combination with an attention mechanism, and high data precision, high consistency and improvement of prediction accuracy are achieved.
Owner:HEFEI OUWO ENVIRONMENTAL PROTECTION TECH CO LTD

Control method and device, storage medium and program product

The invention provides a control method and device, a storage medium and a program product, relates to the technical field of communication, and solves the technical problem of poor beamforming effect of an intelligent metasurface in related technologies. The method comprises the following steps: acquiring a reflection coefficient determination function corresponding to the communication system; the communication system completes communication based on an analog digital hybrid beam, and the reflection coefficient determination function is used for indicating an association relationship between the reflection coefficient of the intelligent metasurface and the spectral efficiency of the communication system; variables in the reflection coefficient determination function are optimized based on an alternating optimization method and a singular value decomposition method, and a target reflection coefficient matrix of the intelligent metasurface corresponding to the maximum spectrum efficiency of the communication system is obtained; and based on the target reflection coefficient matrix, controlling the intelligent metasurface to perform beam forming of the analog-digital hybrid beam.
Owner:CHINA UNITED NETWORK COMM GRP CO LTD

Personalized federal learning method based on singular value decomposition

The invention relates to a singular value decomposition-based personalized federated learning method, which is used for solving the problem of uncertainty of a system caused by equipment isomerism and the problem that a client falls behind due to extra burden caused by a storage and transmission mechanism in the prior art. Firstly, a customer selection strategy based on resource awareness is determined, and a server side establishes a quantitative scoring system by comprehensively analyzing computing resources and communication conditions of clients. And screening a plurality of clients with optimal comprehensive performance from all online clients to participate in local training. During local training of the client, historical weight parameters are reserved by adopting self-knowledge distillation (SKDSVD) combined with a singular value decomposition method, residual errors after decomposition are personalized model parameters and are stored in the local client, and universal parameters are transmitted to the server for aggregation. The SKDSVD algorithm reduces the training time delay and the fall-behind rate of the local client while optimizing the global model of the server to obtain the optimal convergence precision.
Owner:BEIJING UNIV OF TECH

Bolt looseness diagnosis method, system, equipment and medium

The invention relates to a bolt looseness diagnosis method, system and device and a medium. The method comprises the following steps: acquiring an original sound signal for knocking a bolt; decomposing the original sound signal by using a CEEMDAN method to obtain a plurality of IMF components; calculating a plurality of preset types of quantitative indexes for each IMF component, performing dynamic weighting, and screening out a plurality of effective IMF components from the plurality of IMF components based on a weighting result of the IMF components as effective components; determining a dynamic function window and a frequency spectrum gravity center according to the plurality of effective components, decomposing the effective components in combination with a singular value decomposition method, and performing screening and reconstruction to obtain de-noised signals; performing feature extraction on the de-noised signal to obtain a diagnosis feature; and inputting the diagnosis characteristics into a bolt looseness diagnosis model to obtain a bolt looseness diagnosis result. According to the method, the noise reduction effect of the sound signals collected under the working condition can be effectively improved, the obvious bolt looseness diagnosis accuracy and superiority are achieved, and the bolt looseness diagnosis precision and efficiency can be improved.
Owner:HEFEI UNIV OF TECH

Earth surface continuous deformation InSAR monitoring method under vegetation coverage condition

The invention discloses a surface continuous deformation InSAR monitoring method under a vegetation coverage condition. The method comprises the following steps: S1, identifying a pixel unit with a deep monitoring potential; s2, determining and cutting a high-potential region of interest in the range; s3, aggregating and dividing subsets in the time sequence; s4, obtaining an accumulated deformation quantity and an average deformation rate result in a time range of a single time subset based on a matrix singular value decomposition method; and S5, aiming at different scenes, adopting a fusion method of discrete subsets in a time dimension to realize acquisition of an average deformation rate and a high-time-frequency earth surface deformation result of the vegetation-covered earth surface in the whole time sequence. According to the method, limited high-quality phase information is fully utilized, the problem that the coverage degree of an obtained result of a traditional InSAR method is limited is solved, meanwhile, the InSAR operand is reduced, and the application scene of the InSAR technology is expanded.
Owner:CHINA RAILWAY DESIGN GRP CO LTD +2

Large model multi-preference alignment method and device based on hierarchical hybrid expert model

The invention provides a large model multi-preference alignment method and device based on a hierarchical hybrid expert model, and relates to the technical field of natural language processing. The method comprises the following steps: acquiring a pre-trained single-target fine tuning model; extracting a target vector of each single-target strategy in the model, decomposing the target vector through a task vector singular value decomposition method, and generating a low-rank adapter as a LoRA expert of each single target; a PCB-merging and Free-merging fusion model is adopted for processing, and a multi-target LoRA expert is obtained; generating a linear routing layer, and constructing a reward loss function; optimizing the loss function by adopting mirror gradient descent and smooth Chebyshev scaling to obtain a multi-target routing expert; designing a weight router; constructing a hierarchical hybrid expert model according to the multi-target LoRA experts, the multi-target routing experts and the weight routers; and inputting the obtained prompt words and preference vectors input by the user into the hierarchical hybrid expert model, and outputting preferences conforming to the user. According to the invention, the reasoning efficiency can be improved.
Owner:HARBIN INSTITUTE OF TECHNOLOGY (SHENZHEN) (INSTITUTE OF SCIENCE AND TECHNOLOGY INNOVATION HARBIN INSTITUTE OF TECHNOLOGY SHENZHEN)