178 results about "Matrix decomposition" patented technology

The present invention belongs to the technical field of land change detection. Disclosed are a land change detection method combining matrix decomposition and adaptive propagation, and a system. A deep learning semantic segmentation model combining matrix decomposition and adaptive propagation is used for detecting a land change. The model is of a twin double-branch structure, which can receive both an input of a pre-temporal remote sensing image data set and an input of a post-temporal remote sensing image data set and can share network parameters, and which has the advantage of end-to-end feature learning. In the present invention, an attention method based on matrix decomposition is used to replace a conventional attention method based on manual design; moreover, in view of adaptive propagation technology for pixel offsets of neighboring points, a deep learning twin differential semantic segmentation network is established to realize automatic and precise land utilization / land coverage change detection on the basis of high-resolution optical data.

The invention provides an environment monitoring data anomaly detection method, a medium and a system, and belongs to the technical field of electrical digital data process.The method comprises the steps that firstly, an environment monitoring data original matrix is constructed, time sequence segmentation is conducted, and a rule matrix and an anomaly matrix are obtained through matrix decomposition; calculating a historical mean value, a historical standard deviation, a spatial feature and a time feature of the environmental parameters, designing an anomaly detection optimization equation set based on the features, including an environmental parameter weight equation, an anomaly threshold equation and a comprehensive anomaly degree equation, and calculating an anomaly detection result by calculating an environmental parameter weight vector, a dynamic anomaly threshold matrix and a comprehensive anomaly degree index. Finally, dynamic optimization of anomaly detection is achieved through the gain vector and the feedback vector, anomaly detection of real-time environment monitoring data is completed, and the technical problem that in the prior art, an environment monitoring data anomaly detection method is difficult to give consideration to dynamic relevance of multi-dimensional environment parameters and the accuracy of anomaly detection at the same time is solved.

The invention provides a quantum circuit optimization method, device and equipment for a ZUC algorithm subcomponent, and the method comprises the steps: encoding a signal transmitted by communication into quantum bits, and inputting the quantum bits into a ZUC algorithm structure for quantum implementation; representing a first linear subcomponent in the ZUC algorithm structure as a matrix in a binary field; after a matrix in the binary field is expressed as a block matrix, a low-dimensional matrix M in the binary field is derived from the block matrix, and quantum implementation of the M is searched by using a search algorithm designed based on a matrix decomposition principle; and according to a quantum gate parallel rule, carrying out optimized multi-layer division on the quantum of the M again, obtaining the quantum of the M, carrying out optimized layering operation, and optimizing the quantum of the first linear subcomponent. According to the method provided by the invention, the quantum circuit of the subcomponent of the ZUC algorithm is optimized, so that the resource overhead of quantum implementation of the ZUC algorithm is saved.

The invention provides an arch bridge cable force adjusting method and device, a medium and a program product, and relates to the technical field of arch bridge cable force adjusting.The method comprises the steps that by establishing an arch bridge parameterization finite element model, specific unit force is applied to all suspenders to construct a first influence matrix; singular value decomposition is carried out on the first influence matrix, the first k suspenders with the total energy contribution rate exceeding a first preset value are extracted, and the sensitivity of the suspenders to the global suspender cable force is calculated; calculating an importance index in combination with sensitivity and suspender cable force deviation, and screening the first n important key suspenders to construct a second influence matrix; an adjusted vector is obtained based on the difference value of the second influence matrix and the target influence matrix, and the cable force adjustment amount of the key suspender is solved; the key suspender cable force is sequentially adjusted according to the importance sequence. According to the method, the key suspender is accurately identified through matrix decomposition and sensitivity analysis, the adjustment amount is quantified, interaction is considered, full suspender adjustment is avoided, the construction cost is remarkably reduced, and the cable force adjustment efficiency and accuracy are improved.

The invention provides a Cholesky decomposition heterogeneous parallel optimization method and system based on an SW architecture, and relates to the technical field of high-performance computing. The method comprises the following steps: performing sub-block division on a symmetric positive definite matrix based on a distributed parallel distribution scheme, and performing iteration to complete matrix decomposition; each sub-block is distributed to different processes through an MPI programming model, data exchange is carried out between the processes through asynchronous communication, and coarse-grained task-level parallel acceleration is carried out; performing two-stage parallel acceleration on four operations in Cholesky decomposition by utilizing the acceleration parallel characteristic of a master core and a slave core of the SW architecture; wherein for GEMM and SYRK operations, column vectors of a matrix are mapped to a slave core array, and columns are divided according to the number of slave cores; the calculation process is optimized through a double-buffering mechanism, vectorization operation and a loop expansion technology, and the parallel efficiency is improved; and for the TRSM operation, the TRSM operation is decomposed into a plurality of TRSV operations, the TRSV operations are allocated to the slave cores for parallel execution, and a circular reading and data broadcasting mode is adopted to reduce data dependence and realize efficient parallel calculation.

The invention provides a cross-modal semantic alignment method and system and a storage medium. According to the method, multi-modal carbon features are extracted and subjected to standardized preprocessing, the features are mapped to a low-dimensional public semantic space through a projection algorithm, cross-modal semantic association is constructed in combination with matrix decomposition and an association learning model, and finally an alignment result is generated through large model fusion reasoning. The problem that a traditional method is insufficient in precision and poor in semantic coherence in cross-modal semantic alignment is solved.

The invention relates to a multi-target heuristic load flow equation matrix optimal block calculation method and device, and the method comprises the steps: building a Jacobian matrix optimal block model with the minimum load balance and sub-matrix calculation total cost as the target; and taking an objective function of the Jacobian matrix optimal block model as a fitness function of a multi-objective genetic annealing algorithm, and performing Jacobian matrix optimal block calculation to obtain an optimal block division structure of the Jacobian matrix. According to the method, resource consumption during matrix decomposition can be reduced, and the power flow equation matrix decomposition efficiency is improved.

The invention discloses a matrix decomposition and reconstruction method for a large model reasoning scene, and the method comprises the following steps: 1, carrying out the error statistics of conventional nodes of a network layer in a large model, and screening weight nodes capable of carrying out low-rank decomposition through a sensitivity analysis strategy; step 2, gradually performing low-rank setting, decomposing the weight matrixes in the weight nodes which can be subjected to low-rank decomposition and are obtained by screening in the step 1, and obtaining and recording respective optimal low-rank values of the decomposable weight matrixes; and step 3, performing network reconstruction on the original model based on the decomposable weight matrix and the corresponding low-rank value. The invention further discloses a system for implementing the method and a corresponding application, and the system and the application have a wide application scene.

The invention discloses a multi-point non-stationary non-Gaussian random process efficient simulation method based on a spectral representation method, and belongs to the technical field of random process simulation. Aiming at the problem of low efficiency of existing potential Gaussian random process power spectrum matrix solving and matrix decomposition, the method comprises the following steps: determining a power spectrum matrix and non-Gaussian features of a target process, establishing a transfer equation by using a Mehler formula, calculating time and frequency interpolation points based on the amplitude of the power spectrum matrix, and performing matrix decomposition and orthogonal decoupling at the interpolation points. And generating a potential Gaussian process sample through interpolation and fast Fourier transform, and finally converting to obtain a target process sample. According to the method, time-frequency domain resource allocation is optimized through a self-adaptive interpolation strategy, so that interpolation points are concentrated in an area with high energy, redundancy decomposition is avoided, and the calculation efficiency is remarkably improved on the premise of ensuring the simulation precision.

The invention discloses a hyperspectral fast imaging method and system based on spectral difference depth matrix decomposition, and aims at low signal-to-noise ratio data under a limit imaging condition to realize high-quality image recovery through structural modeling of an average spectrum chart and a spectral difference chart. The method comprises the following steps: firstly, quickly collecting and splicing hyperspectral data to form a three-dimensional tensor, and constructing a forward observation model under a weak light condition; decomposing into an average spectrum chart, a spectrum difference chart and a noise item; rearranging the spectral difference graph into a two-dimensional low-rank matrix, and obtaining a space and spectral principal component matrix by adopting PCA decomposition; an unsupervised dual-path optimization network is innovatively constructed; a 2D jump network is used to optimize a spatial principal component so as to model structural information, and a 1D jump network is used to optimize a spectral principal component so as to fit detail changes; and finally, outputting an optimized spectrum difference graph by the network after joint training, and fusing the optimized spectrum difference graph with the average spectrum graph to complete reconstruction. High-fidelity image reconstruction is realized under single-frame and unsupervised conditions, and the reconstruction quality and robustness are significantly improved.

The invention discloses an overlapping community detection method, system and device and a medium, and particularly relates to the technical field of community detection, and the technical key points are as follows: extracting node information data from a pre-constructed adjacent matrix of an overlapping community, and inputting the node information data into a point mutual information function to calculate and obtain point mutual information; obtaining a superpoint mutual information matrix by combining the point mutual information with the hypergraph structure; constructing a target function by using the super-point mutual information matrix and a pre-constructed three-factor illegal matrix decomposition optimization model based on graph regularization; decomposing the super-point mutual information matrix in the objective function to obtain an indication matrix, and solving the indication matrix by using an alternating iteration method to obtain an optimal solution of the indication matrix; and detecting the overlapping community based on the optimal solution of the indication matrix to obtain an overlapping community detection result.

The invention discloses a multi-signal-source scene-oriented generative radio map collaborative estimation method, which comprises the following steps of: S1, sampling received signal strength of a plurality of users on a plurality of frequencies to construct a matrix form, and decomposing by adopting an NMF matrix decomposition method to obtain an NMF matrix; each decomposed received signal strength component corresponds to an independent signal source; s2, constructing and training a radio map inference model of a single signal source; and S3, for each received signal strength component obtained by decomposition in the step S1, estimating correspondence by using a radio map reasoning model of a single signal source. According to the method, the signal intensity components corresponding to multiple signal sources are decomposed by adopting the matrix decomposition theory, and the neural network is trained by utilizing the learning algorithm based on the GAN, so that the accuracy and generalization ability of the system are improved. The method can effectively work under the condition that geographic data is inaccurate or matrix decomposition results have errors.

The invention relates to a power network structure weakness detection method based on multi-diagonal-block matrix decomposition, and belongs to the technical field of super-large-scale integrated circuits. According to the method, an original large-scale sparse matrix is converted into a band edge diagonal block structure and divided into a plurality of sub-matrixes capable of being solved independently by constructing a layering and blocking strategy for eliminating tree drive, and redundancy of full-matrix calculation is avoided. Meanwhile, a local approximate inverse algorithm of column norm truncation is designed, and target elements are calculated on the premise that a preset error threshold value is met. According to the method, the parallel computing architecture of the multi-core processor is fully utilized, the computing complexity is reduced by 2-3 orders of magnitude, the computing efficiency is effectively improved, the memory occupation is remarkably reduced, and a high-precision and high-efficiency solution is provided for detecting the weakness of the power network structure of the super-large-scale integrated circuit.

The invention provides a target recognition method and device based on a low-rank neural network, and relates to the field of radar target recognition. According to the scheme, CP decomposition is carried out on a convolution kernel weight tensor of a two-dimensional convolution kernel in a convolution module to obtain a low-rank convolution module; the decomposition rank of CP decomposition is determined according to the Kruskal uniqueness theory and the computing power of low-rank neural network operation equipment; performing matrix decomposition on the full-connection layer weight matrix of the full-connection module to obtain a low-rank full-connection module; the decomposition rank of matrix decomposition is determined according to the input feature dimension and the output feature dimension of the full-connection module; replacing a convolution module and a full connection module in the convolutional neural network with a low-rank convolution module and a low-rank full connection module to obtain a low-rank neural network; and collecting target feature parameters, inputting the target feature parameters into the low-rank neural network, and obtaining a target recognition result. By using the method, relatively small performance loss can be kept under the condition that the model parameter quantity is greatly compressed.

The invention provides an artificial intelligence model accelerated training reasoning method and device, and belongs to the field of artificial intelligence, and the method comprises the steps: firstly carrying out the grouping disassembly of a large-dimension matrix in an artificial intelligence model through employing a lamination grouping technology, and converting the large-dimension matrix into a lamination block matrix structure; then, carrying out compression decomposition on the non-diagonal block matrix by utilizing an ACA-SVD matrix decomposition technology, and representing that two small matrixes are multiplied; then diagonal matrix blocks of the obtained laminated block matrix structure are extracted in sequence, the matrix is expressed as multiplication of a plurality of block diagonal matrixes, and the block diagonal matrixes are stored respectively; when model training and reasoning are carried out, a matrix multiplication process can be converted into multiplication of a plurality of block diagonal matrixes, a matrix inversion process can be converted into inversion of the plurality of block diagonal matrixes, the block diagonal matrixes are mutually independent, a recursive relationship does not exist in an operation process, and highly parallel operation can be realized.

The invention relates to the technical field of data processing, in particular to a dynamic weight matrix decomposition energy consumption prediction method and system, and the method comprises the steps: standardizing internal energy consumption data, and generating a standard data matrix and a sparse mask matrix; based on the mask matrix, processing the standard data matrix by adopting a dynamic regularization matrix decomposition technology, and determining a sparse internal factor matrix and a reconstructed data matrix; and smoothing the noise points by using a local weighted regression algorithm to generate a smooth data matrix. Standardizing the external influence factor data to form low-dimensional external factor embedding; and calculating dynamic influence weights of the external factors by adopting an attention mechanism algorithm, and fusing to generate weighted external factor representation. And when newly added energy consumption data is obtained, a final factor matrix is obtained through an incremental learning optimization algorithm. And outputting a future energy consumption prediction result based on the final factor matrix and the smooth data matrix, thereby effectively improving the accuracy, robustness and efficiency of energy consumption prediction in a complex scene.

The invention discloses a vehicle spatial position uncertainty modeling method based on a factor graph, and relates to the technical field of automatic driving. Establishing a factor graph model, and constructing state variable nodes and factor nodes; an objective function of the factor graph optimization problem is converted into a nonlinear least square problem, and optimal state estimation is obtained through iterative solution by adopting a Gauss-Newton method; on the basis of an objective function under a Gaussian Newton method framework, a posterior covariance matrix is obtained through calculation of a Hesse matrix and inversion; performing incremental optimization on the factor graph, and updating a global Jacobian matrix through matrix decomposition and Givens transformation; and carrying out confidence ellipse geometric quantization, and representing the positioning uncertainty in a geometric form. Based on a factor graph and a confidence ellipse theory, through multi-sensor space-time constraint fusion and geometric interpretable representation, a vehicle positioning uncertainty modeling framework is constructed, and the safety and robustness of an automatic driving system downstream decision planning algorithm are improved.

The invention discloses a multi-mode private illegal content retrieval method and system based on semantic granularity perception, and belongs to the technical field of image-text multi-mode content retrieval. In order to solve the technical problem that private and illegal contents are difficult to effectively retrieve through explicit text description, the method comprises the following steps of: constructing an image identifier consisting of clustering, matrix decomposition, quantization and unique coding, and realizing generative learning of an image and the image identifier by utilizing a multi-modal large model; and constructing a soft label based on the similarity between the image and the text representation, and outputting a corresponding image identifier through text-driven generative retrieval learning. According to the method, the illegal image content with semantic matching can be retrieved from the private text description, and the method is suitable for scenes such as illegal content examination, risk identification and content filtering.

The invention relates to the technical field of data feature extraction, in particular to a high-dimensional data feature extraction and dimension reduction processing system and method based on matrix factorization, and the system comprises a data acquisition module which is used for reading an original high-dimensional matrix; the preprocessing module is used for receiving the original high-dimensional matrix and outputting a preprocessed matrix; the noise pattern recognition module is used for receiving the preprocessed matrix and outputting a noise category label vector and a confidence coefficient vector; the weight matrix generation module is used for receiving the noise category label vector and the confidence coefficient vector and outputting a diagonal weight matrix; the matrix decomposition module is used for receiving the preprocessed matrix and the diagonal weight matrix and outputting a low-dimensional feature matrix and a basis matrix; and the post-processing module is used for receiving the low-dimensional feature matrix and outputting a normalized feature matrix. According to the method, the discrimination and decomposition precision of low-dimensional features are remarkably improved through self-adaptive optimization of emphatically reserving effective samples and suppressing noise sample interference.

The present invention relates to the field of product evaluation, and more specifically, to a product quality evaluation method and system based on big data. The method comprises: collecting multi-dimensional detection data, obtaining a quality evaluation value of a product, and calculating the degree of abnormality; calculating the degree of influence of the detection data of a target dimension on a target product; respectively calculating the correlation between the detection data of the target dimension and the detection data of each dimension, and using the global influence degree to correct the correlation coefficient correction value, constructing a correlation coefficient correction value matrix, and using matrix decomposition to obtain the weight of the detection data of the target dimension during quality evaluation; respectively multiplying the weight by the detection data of the corresponding dimension to obtain new multi-dimensional detection data to complete product quality evaluation. Through the technical solution of the present invention, the accuracy and efficiency of product quality evaluation can be improved.

A method for compressing a plaintext matrix and a computing device, the plaintext matrix is obtained, the plaintext matrix is a Fourier transform matrix or an inverse Fourier transform matrix, the plaintext matrix is decomposed to obtain a plurality of sparse matrixes, the number of non-zero diagonals of each sparse matrix except a main diagonal is at most 2, and the product of the plurality of sparse matrixes is the plaintext matrix, dividing the plurality of sparse matrixes into n groups, aiming at any group of sparse matrixes, determining a product of the group of sparse matrixes as a combined matrix corresponding to the group of sparse matrixes, aiming at any combined matrix, determining an access period corresponding to the combined matrix, and respectively performing access on the combined matrix along each diagonal line according to the access period, the compression diagonals corresponding to the combined matrix are determined according to the access result, and the plaintext matrix can be compressed by utilizing the repetition rule of the diagonal elements of the combined matrix obtained after the plaintext matrix is decomposed and combined, so that the space required for storing the plaintext matrix can be reduced, and the efficiency of transmitting the plaintext matrix can be improved.

The present application relates to a kind of multi-objective heuristic power flow equation matrix optimal block computing method and device, wherein, method includes: with load balance and sub-matrix total cost minimum as target, construct Jacobian matrix optimal block model;With the objective function of the Jacobian matrix optimal block model as the fitness function of multi-objective genetic annealing algorithm, and carry out Jacobian matrix optimal block calculation, obtain the optimal block division structure of Jacobian matrix.The present application can reduce the resource consumption when matrix decomposition, accelerate the efficiency of power flow equation matrix decomposition.

The invention discloses an expressway traffic flow missing data complementing method based on space-time multi-scale matrix factorization, which comprises the following steps: acquiring expressway multi-source heterogeneous traffic data, and constructing an observation matrix, wherein N represents the number of spatial positions, and T represents the number of time steps; decomposing the observation matrix into a product of a multi-scale space factor matrix and a time factor matrix by adopting Bayesian time sequence matrix decomposition; splitting the multi-scale space factor matrix into road section-level, region-level and network-level feature matrixes, and introducing dynamic evolution of a multi-order vector autoregression process modeling time factor matrix; performing model parameter learning through a Bayesian inference framework and a Gibbs sampling algorithm; and reconstructing an observation matrix by using W and X obtained by Bayesian inference, and outputting a complemented data matrix. According to the method, the problems of weak expression ability, high calculation complexity, limited missing value processing ability and the like of a traditional interpolation method are effectively solved, and the data completion precision is remarkably improved.

The invention provides a risk early warning method based on blast-furnace tuyere image data processing, a medium and a system, and belongs to the technical field of image data processing.The risk early warning method comprises the steps that firstly, video data of a blast-furnace tuyere are collected, and an image sequence is extracted according to a preset time interval; in order to eliminate the brightness difference caused by the environment illumination change, the image sequence is subjected to brightness equalization processing. Then, the color space is converted from RGB to HSV, and a color feature matrix is extracted. And through clustering analysis, the slag adhering state and the large block falling state of the blast furnace tuyere area are recognized. In addition, a time sequence of a color feature matrix is established, a stable component and a fluctuation component are extracted by adopting a matrix decomposition method, and the fluctuation component represents a state change trend. And after dimensionality reduction of the change trend matrix, inputting the change trend matrix into a preset risk discrimination matrix equation set to obtain a slag adhering risk index and a large block falling risk index. And finally, performing threshold analysis according to the risk vector, outputting a final risk result, and realizing risk early warning of slag adhering and large block falling.

The invention relates to the technical field of data processing, and discloses a large matrix characteristic decomposition method, system, storage medium and device, the method carries out matrix decomposition based on House holder transformation and Givens transformation, and the problems of low efficiency and the like in the prior art are solved.

The invention discloses a variable-order covariance matrix inversion method, and belongs to the technical field of array signal processing. Aiming at the inversion problem of the covariance matrix, according to the conjugate symmetry property of the covariance matrix, the covariance matrix inversion operation based on the FPGA is realized by utilizing a cholesky matrix decomposition method. The FPGA implementation architecture comprises a state control module, a state machine module, a storage module and a calculation module. According to the method, in a parameterization design mode, a user can customize the order number and the data bit width of an inversion matrix, the use of logic resources of an FPGA is saved in a time division multiplexing mode, and the operation speed is increased in an address index pipeline processing mode. According to the design, the self-definition of a user, the utilization rate of resources and the calculation precision and time are integrated, good compatibility is achieved, and the universal requirement for covariance matrix inversion is met.

Implementations disclosed describe techniques used for compiling a quantum algorithm for execution on a plurality of quantum circuits, including accessing, by a processing device, the quantum algorithm, identifying a matrix associated with the quantum algorithm, determining a representation of the identified matrix as a matrix decomposition that includes a plurality of transformation matrices, wherein one or more of the plurality of transformation matrices perform multiple instances of two-dimensional rotations; and generating a circuit map that maps execution of the matrix decomposition on the plurality of quantum circuits.

The invention discloses a series fault arc identification method and system based on singular spectrum statistical characteristics, and the method comprises the steps: S1, collecting the current waveform of a series AC fault arc, and obtaining a current sampling sequence; s2, constructing a Hankel matrix Y based on the current sampling sequence, and then performing singular value decomposition on the Hankel matrix Y to obtain a singular spectrum; s3, statistical features are calculated based on singular spectrums, and feature vectors are formed; and S4, inputting the feature vector into a trained XGBoost classifier to obtain a category label of the series AC fault arc, and realizing real-time identification of the fault arc. According to the method, the singular spectrum of the current waveform is constructed through Hankel matrix decomposition, and the high-frequency harmonic and amplitude characteristics of the fault arc are efficiently captured through the singular spectrum, so that the recognition precision is improved; and the hyper-parameters of the classifier are optimized in combination with a differential evolution algorithm, so that the recognition efficiency of the series alternating-current fault arc is improved.

The invention discloses a high-efficiency low-rank tensor recovery method. The method comprises the following steps: constructing an equilibrium tensor; a reconstruction operation is given for an unbalanced tensor, and a more balanced tensor is obtained; a low-rank tensor recovery algorithm based on a non-convex Logdet function is constructed; a non-convex Logdet function is used for replacing a nuclear norm, and a better approximation matrix rank is achieved; introducing a matrix decomposition method; through matrix decomposition, a large-scale matrix is decomposed into a product of two small matrixes, and the size of the matrix is reduced; fusing the above methods to obtain an efficient low-rank tensor recovery algorithm; according to the scheme, the equilibrium tensor and the non-convex function are fused, so that high-precision low-rank tensor recovery is realized, and furthermore, the efficiency of a low-rank tensor recovery algorithm can be effectively improved by introducing a matrix decomposition method.

The invention provides a multi-modal model training method and device. The method comprises the following steps: training each decomposition matrix in a multi-modal model by using training data until the loss value of a preset loss function is smaller than a first preset threshold value, and obtaining a trained multi-modal model; wherein the parameter matrix of the multi-modal model comprises a basic matrix and an increment matrix of each layer, the decomposition matrix is obtained by decomposing the increment matrix, one decomposition matrix corresponding to the increment matrix is a same-order matrix, other decomposition matrixes are non-same-order matrixes, and the total parameter quantity of the decomposition matrixes is smaller than the parameter quantity of the increment matrix. According to the method, the decomposition matrix with a small parameter quantity in the multi-modal model is trained, the fine tuning training time is shortened, the training efficiency of the model is improved, the demand of hardware video memory resources is reduced, the training cost is reduced, the precision improvement of downstream tasks of the multi-modal model is completed by using a small amount of video memory resources and training time, and the training efficiency of the multi-modal model is improved. And the method has relatively strong generalization ability.