58 results about "Matrix decomposition" patented technology

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 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.

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.

This invention discloses an efficient QR matrix decomposition method based on FPGA for implementing systolic matrices, belonging to the field of software-defined radio signal transmission technology for airborne networks. It is based on the design of a multi-channel digital front-end implemented on an FPGA chip. The method includes: designing a general handshake coordination mechanism, designing a timing flow for systolic matrix multiplication, performing fast QR matrix decomposition single-loop operations by calculating the Householder mirror transformation matrix, designing a top-level loop pipeline for QR matrix decomposition, and finally verifying the hardware scheme of the fast QR decomposition algorithm. This invention converts resource consumption into timing consumption by designing the overall pipeline logic and loop operations, enabling fast processing and accurate decomposition of digital information in a multi-channel digital front-end on an FPGA, effectively saving on-chip resources. In addition, this invention has excellent scalability, supporting 64 to 128-bit high-precision fixed-point number formats, and the data matrix order can reach up to 128.

The application discloses a kind of recommendation algorithm based on label neighborhood model and label matrix decomposition model, it is related to user recommendation technical field, including the construction UTagUser-CF algorithm and TagMF algorithm, according to the UTagUser-CF algorithm and TagMF algorithm constructed to construct UTagUser-TagMF algorithm;The recommendation algorithm based on label neighborhood model and label matrix decomposition model, by using TagMF algorithm, so that the algorithm has lower recommendation error, better recommendation effect, using label information model helps matrix decomposition model to increase implicit feature information, to further solve the cold start problem and data sparsity problem, to solve the accuracy problem, using UTagUser-TagMF algorithm when facing the cold start problem, so that the algorithm has better recommendation performance, can effectively solve the cold start problem and data sparsity problem.

This invention relates to a direction-finding method based on matrix decomposition, belonging to the field of array signal processing technology. The invention first calculates the covariance matrix of the received data, then performs matrix decomposition and reconstruction based on the covariance matrix, and finally constructs a solution function by combining the spatial relationship between the reconstructed matrix and the array steering vector. The direction of arrival of the signal source is obtained from the mapping relationship between the peak value of the solution function and the signal incident angle. This invention utilizes the orthogonality between noise vectors to reduce the impact of noise on the direction of arrival (DOA), which can further improve the stability of the direction-finding algorithm.

The invention discloses an ellipsometry depolarization correction method and system, and the method comprises the steps: measuring a to-be-measured sample through a Mueller matrix ellipsometer, and obtaining a measurement Mueller matrix Me of the to-be-measured sample; decomposing the measurement Mueller matrix Me to determine a depolarization matrix M3 representing a depolarization effect; and correcting the measurement Mueller matrix Me by using the depolarization matrix M3 to obtain a corrected Mueller matrix M0. According to the method, the depolarization effect can be effectively separated and corrected through the accurate matrix decomposition and correction process, and interference of the depolarization effect on the measurement result is avoided. The innovative nonlinear fitting method combines the corrected Mueller matrix and depolarization matrix, and constructs a joint evaluation function for global optimization, so that the extraction of optical parameters is more accurate and physically self-consistent. The iterative optimization step further improves the measurement precision, and compared with the prior art, the measurement problem of a complex sample can be better solved, and a more reliable solution is provided for the field of optical measurement.

The present application relates to the technical field of computer vision, and relates to a 3D Gaussian sputtering stylization method, system, device and storage medium; wherein the 3D Gaussian sputtering stylization method comprises the following steps: obtaining a style image and a three-dimensional Gaussian scene to be stylized; performing non-negative matrix decomposition on the style image to extract an initial color basis matrix and an initial coefficient matrix of the style image; constructing a color orthogonal decoding model according to the initial color basis matrix and the initial coefficient matrix, and recalculating color parameters of each Gaussian primitive of the three-dimensional Gaussian scene; performing style transfer optimization on the Gaussian primitive, updating trainable parameters until an iteration termination condition is reached; and determining a stylized three-dimensional Gaussian scene according to updated geometric attribute parameters and color attribute parameters. The present application can solve the problems of color incoordination and dirty color during 3D Gaussian sputtering stylization.

The invention relates to the technical field of carbon emission calculation, and discloses an electric power carbon emission factor calculation method, system, equipment, medium and product based on QR matrix factorization. According to the method, a node power balance equation and a carbon emission conservation equation are determined by obtaining carbon emission information and tidal current power information of all generators in a power grid; the method comprises the following steps: constructing a carbon emission factor linear equation, solving the carbon emission factor linear equation based on QR matrix decomposition to obtain a carbon emission factor value of each node, so that efficient solving is realized by utilizing sparse matrix characteristics and QR decomposition, rapid response of multi-node carbon emission factor solving is realized, and the method does not depend on initial solution setting; the slow convergence or failure risk of an iteration method is avoided, the numerical stability is excellent, and the accuracy and reliability of carbon emission factor calculation are improved.

The application discloses a kind of based on gradient descent method's uniform array ultra-wideband radar positioning imaging method, it is related to radar positioning technical field, the signal matrix model after matching filtering is constructed in the application, matrix decomposition is carried out by gradient descent method, joint direction matrix and joint electromagnetic guide vector matrix are obtained, then, two groups of two-dimensional angle estimation values with high resolution and no ambiguity are respectively obtained from the two matrices, accurate angle parameters are obtained by weighted fusion, and target three-dimensional coordinates are calculated accordingly, finally, radar image is generated by back projection algorithm, so that the number of array elements and hardware complexity are significantly reduced, the inherent angle ambiguity problem of array is overcome, the multi-target angle resolution and parameter estimation stability are improved, and the calculation efficiency and engineering realizability are also considered.

The application discloses a project recommendation method and device, a computer readable storage medium and an electronic device. It relates to the field of recommendation algorithm, and the method comprises the following steps: obtaining the scores of each user in N users to at least part of M projects, and constructing a score matrix according to the scores; randomly generating an initial user feature matrix, an initial hidden vector matrix and an initial project feature matrix; updating the initial user feature matrix, the initial hidden vector matrix and the initial project feature matrix based on a matrix decomposition method until the iteration of the matrix decomposition method meets a preset iteration condition; determining the predicted scores of each user to all the M projects according to the user feature matrix, the hidden vector matrix and the project feature matrix; determining the to-be-recommended projects corresponding to the N users respectively according to the predicted scores; and recommending the to-be-recommended projects to the N users. The application solves the technical problem of low recommendation accuracy when recommending projects to users based on matrix decomposition in the related art.

The invention discloses an AUV (Autonomous Underwater Vehicle) velocity field reduction method based on intrinsic orthogonal decomposition and neural network fusion, which comprises the following steps: constructing a velocity field snapshot matrix of an AUV, each velocity field snapshot being a multi-dimensional velocity field obtained by training sample simulation; preprocessing the speed field snapshot matrix, and then performing matrix decomposition by using a POD method to obtain a standard orthogonal basis matrix and a diagonal matrix; setting a truncation coefficient, determining the number of singular values needing to be reserved from the diagonal matrix by using the truncation coefficient, performing dimensionality reduction on the standard orthogonal basis matrix by taking the number as a dimensionality reduction order, and determining an order reduction coefficient matrix based on the standard orthogonal basis matrix after dimensionality reduction; constructing a working condition matrix, wherein the working condition matrix comprises working conditions of different dimensions; and fitting a mapping relation between the working condition matrix and the reduced-order coefficient transpose matrix by using a neural network so as to predict the velocity field under the to-be-predicted working condition.

The application discloses a reply text generation method and device, and relates to the technical field of model optimization, which comprises the following steps: performing quantized tensor column decomposition on a feature training matrix in a model training process to calculate a weight matrix and a bias; in a pre-filling stage of model inference, performing matrix multiplication and addition operation on the feature matrix after the matrix tensor decomposition operation by using the weight matrix and the bias in the form of tensor decomposition after the model training; in a decoding stage, performing tensor decomposition on a one-dimensional feature vector to perform matrix multiplication and addition with the weight matrix and the bias in the form of tensor decomposition, so as to generate a target reply text based on the corresponding calculation result, thereby solving the technical problem that the selection requirement of the rank in the matrix decomposition calculation process is high in the related art, the complexity is difficult to effectively reduce, or the error between the calculation result obtained by using the decomposition form and the original result is large, and the technical effect of greatly reducing the storage and calculation complexity and effectively improving the inference performance is achieved.

The invention provides an optical calculation method, an optical calculation system, a storage medium and a program product. The optical calculation method comprises the following steps: acquiring an input matrix and an input optical signal; wherein the input matrix is a non-unitary matrix; performing matrix decomposition processing and inverse calculation processing on the input matrix to generate an expansion matrix corresponding to the input matrix; wherein the expansion matrix is composed of an input matrix, a matrix subjected to matrix decomposition processing and a matrix subjected to inverse calculation processing; providing an optical neural network, calculating a programming phase value of the expansion matrix in each calculation unit of the optical neural network, and programming the programming phase values to the optical neural network; and inputting an input optical signal into the programmed optical neural network, and generating and outputting an input matrix and an input optical signal calculation result. Therefore, multiplication of the non-unitary matrix can be realized, and the universality of the optical neural network is improved.

The application discloses a matrix decomposition method based on secure aggregation and key exchange, and provides a new idea for enhancing data security of federated learning by performing secure aggregation on the gradient of an item matrix I of matrix decomposition under a federated learning framework; the training sample of a recommendation model (namely, a federated learning model) is efficiently utilized by using the local and securely aggregated gradient, so that the user data is ensured not to leave the local, and meanwhile, the recommendation model training process is made more secure; the gradient is masked and added with noise, so that the leakage of source data information caused by exposure of the real gradient is effectively avoided; and the gradient aggregation mode based on secure aggregation is provided, and compared with the homomorphic encryption technology adopted in the background art, the gradient encryption and decryption have lower calculation complexity and faster calculation speed, and the training speed of the recommendation model is improved.

The present application relates to a kind of optimal column ordering method and device of tide Jacobian matrix, wherein, method includes: the maximum non-zero element number of optimal pivot column in lower triangular matrix in Jacobian matrix decomposition process is calculated, and the non-zero element upper bound of optimal pivot column is obtained;With the same non-zero element structure of optimal pivot column, pivot row mode skip list and original row mode skip list are updated to empty set;The numerical value of each column number value corresponding to row mode counter in the non-zero element upper bound mode skip list of pivot row of the column number value pointed to by the column number value of optimal pivot column is updated, and the numerical value of pivot row mode counter and original row mode counter is obtained;The column metric value of each column number value in the non-zero element upper bound mode skip list of pivot row of the column number value pointed to by the column number value of optimal pivot column is updated and calculated.The present application can improve the response speed and effectiveness of power flow calculation.

The application relates to a user attribute inference method and device based on hierarchical multi-channel hypergraph modeling, wherein the method comprises the following steps: constructing hierarchical multi-channel attribute hyperedge groups and interaction hyperedge groups of users and items based on network data; constructing hypergraph pairs based on hierarchical multi-channel attribute hyperedge groups and interaction hyperedge groups fusion based on an attention mechanism; constructing a hierarchical multi-channel hypergraph convolution network according to the hypergraph pairs and corresponding hypergraph association matrices, and learning hypergraph embedding; and calculating user attributes according to overall representation and matrix decomposition, and optimizing a collaborative filtering task and a user attribute inference task. Thus, the problems in the related art that the collaborative filtering and the user attribute modeling inference task are not alternately optimized and mutually enhanced, the association in the user-item interaction network with attribute data is complex, it is difficult to accurately model high-order complex association therein, the attribute data and the interaction data are relatively sparse, and the user attribute inference effect is reduced are solved.

The application discloses a meshless frequency estimation method for layout unconstrained blade tip timing signal and belongs to the technical field of array signal processing. The method comprises the following steps: setting a sampling model; setting the sampling model of the irregular layout of the blade tip timing signal according to the basic principle of the blade tip timing and the continuous compressive sensing theory; obtaining the rank constraint optimization problem description based on the atomic norm from the sampling model; obtaining the characteristic matrix through the irregular Vandermonde matrix decomposition; solving the rank constraint optimization problem based on the atomic norm through the alternating projection algorithm; and obtaining the frequency estimation value of the signal from the irregular Vandermonde matrix based on the polynomial root estimation method of the root multiple signal classification. The application is based on the basic principle of the blade tip timing and the continuous compressive sensing theory, and the frequency grid does not need to be divided in advance, so that the sensor layout is not restricted by the sparse linear array layout, and the blade tip timing signal obtained is uniform.

The invention discloses a reply text generation method and device, and relates to the technical field of model optimization, and the method comprises the steps: carrying out the quantization tensor column decomposition of a feature training matrix in a model training process, and carrying out the calculation of a weight matrix and bias; in a pre-filling stage of model reasoning, carrying out matrix multiplication and addition operation by utilizing a weight matrix and bias in a tensor decomposition form after model training and a feature matrix after matrix tensor decomposition operation; in the decoding stage, tensor decomposition is carried out on the one-dimensional feature vector, matrix multiplication and addition are carried out through a weight matrix and bias in a tensor decomposition form, and a target reply text is generated based on a corresponding calculation result, so that the problems that in the related technology, the selection requirement of a rank in the matrix decomposition calculation process is high, the complexity is difficult to effectively reduce, and the decoding efficiency is high are solved. Or the error between the calculation result obtained by utilizing the decomposition form and the original result is large is solved, and the technical effects of greatly reducing the storage and calculation complexity and effectively improving the reasoning performance are achieved.

A data compression method, a data compression system and an operation method of a deep learning acceleration chip are provided. The data compression method includes the following steps. A filter coefficient tensor matrix of a deep learning model is obtained. A matrix decomposition procedure is performed according to the filter coefficient tensor matrix to obtain a sparse tensor matrix and a transformation matrix, which is an orthonormal matrix. The product of the transformation matrix and the filter coefficient tensor matrix is the sparse tensor matrix. The sparse tensor matrix is compressed. The sparse tensor matrix and the transformation matrix, or the sparse tensor matrix and a restoration matrix, are stored in a memory. A convolution operation result is obtained by the deep learning acceleration chip using the sparse tensor matrix. The convolution operation result is restored by the deep learning acceleration chip using the restoration matrix.

The invention relates to the technical field of multi-label feature selection, in particular to a sparse multi-label feature selection method based on feature positions, which comprises the following steps of: decomposing an original feature space matrix by using a coupling matrix decomposition method; the basis matrix alignment of the feature space and the label space is realized through dynamic graph Laplacian; alternately optimizing to obtain a global optimal coefficient matrix; fusing information entropy and coefficient matrix score weighting to enhance double spaces; constructing an objective function through norms; and optimizing the objective function based on an alternating direction multiplier method and a norm threshold method to obtain a globally optimal solution. According to the method, a double-space interaction matrix is obtained through composite matrix decomposition and a dynamic graph Laplacian technology, interaction features and label weights are obtained through F-norm, and index normalization weighting is performed on the interaction weights by introducing information entropies of an original feature space and a label space, so that the label space and the feature space are enhanced; and high-dimensional structures among the data are captured to obtain valuable features.

This invention discloses a Bayesian inversion extraction method for the equilibrium signal of wide-span altimeter data. The method involves acquiring and preprocessing sea surface height anomaly sequences from radar interferometers and nadir altimeters. A normalized sine square window function is applied for windowing, and multidimensional spatial averaging is used to estimate the one-dimensional wavenumber power spectrum. A piecewise power law-based equilibrium signal spectrum model and a noise spectrum model constrained by dynamic sea state are constructed, and the set of spectral parameters is extracted through logarithmic domain weighted least squares fitting. A set of spatial covariance matrices is constructed using cosine integral transform and Abelian forward and inverse transforms. A graphics processor is scheduled to perform batch matrix decomposition and singular fault-tolerant regularized inversion to solve for the posterior mean vector and posterior covariance matrix of the target equilibrium signal. Window fusion and index mapping are applied to fill the gaps in nadir observations. Geostrophic dynamics parameters are calculated, uncertainty quantification is performed based on the linear error propagation law, and the knowledge base is updated based on the exponential moving average algorithm. This invention achieves suppression of observation noise and physical filling of observation gaps, improving the adaptability of the inversion system to environmental changes while preserving non-Gaussian dynamic characteristics.

The embodiment of the invention relates to a data processing method and device, equipment, a medium and a program. Input data can be acquired; processing the input data through a low-rank adaptive model to obtain output data corresponding to the input data; the acquisition process of the low-rank adaptive model comprises the following steps: determining a first matrix of a first attention layer in a fine tuning model; determining a second matrix of a second attention layer in a base model corresponding to the fine tuning model; determining a first increment matrix of the first matrix relative to the second matrix; matrix decomposition is carried out on the first increment matrix, a first decomposition matrix and a second decomposition matrix of the first increment matrix are obtained, and the rank of the first decomposition matrix and the rank of the second decomposition matrix are smaller than the rank of the first increment matrix; and generating a low-rank adaptive model based on the first decomposition matrix, the second decomposition matrix and a third matrix of the first attention layer. Therefore, the balance between the reasoning accuracy and efficiency of the low-rank adaptive model can be better realized.

The invention provides a federated learning course recommendation method based on flying paddle PadlePaddle, and the method achieves the safety training and efficient deployment of a multi-college collaborative recommendation model through the construction of a cloud-edge-end three-stage architecture and the fusion of matrix decomposition modeling, a federated learning mechanism and a differential privacy protection strategy. The specific method comprises the following steps: completing data preprocessing and coding at an edge node, and constructing low-dimensional representation of a user and a course by utilizing a double-embedded matrix; in the local training stage, a joint loss function optimization model is combined, and differential privacy protection is realized through gradient cutting and Gaussian noise injection; the cloud end aggregates each node model through FedAvg and issues an update; and finally, a Top-K personalized recommendation result is generated and output in a structured manner, and a butt joint teaching system is supported. The federated learning course recommendation method solves the problems that data islands exist in the education field and privacy protection and recommendation precision are difficult to consider at the same time, validity is verified in a MOOCCube college data set, data privacy can be ensured, and high recommendation accuracy is kept.

The application provides a kind of space-time crack data inference method based on graph and attention, comprising the following steps;Step S1: obtain the original sparse matrix containing space-time crack data;Step S2: remove all space-time cracks in the original sparse matrix to obtain a smaller size crack-free sparse matrix;Step S3: infer the crack-free sparse matrix using matrix decomposition techniques;Step S4: put back the space-time cracks removed in step S2 to obtain a preprocessed sparse matrix;Step S5: construct a spatial block based on a graph attention network and a time block based on a Transformer, and combine them into a spatio-temporal neural network structure;Step S6: input the preprocessed sparse matrix into the model of the spatio-temporal neural network structure, obtain the inferred matrix and calculate the loss;Step S7: repeat step S6 to train the model until convergence;The application can effectively infer more ideal complete data given perception data with sparsity and possibly accompanied by space-time crack characteristics.

This invention discloses a method and system for modeling the dynamics of a five-axis machine tool based on monoid coding. The method includes: S1, deriving the modular kinematics of a cascaded five-axis machine tool based on local exponential products; S2, completing the linearization modeling of the modular kinematics with respect to the kinematic parameters; S3, obtaining the linear representations of the inertia, centrifugal force, Coriolis force, and gravity terms of the dynamic model with respect to a multivariate polynomial; S4, describing the linearly represented dynamic model as a bilinear representation with respect to the inertia parameters and the multivariate polynomial; S5, eliminating redundant multivariate polynomials using matrix row correlation and eliminating redundant parameters using matrix decomposition to obtain the minimum set of inertia parameters; S6, inversely encoding the multivariate polynomial encoding numerical matrix and the parameter matrix after redundancy elimination into a symbolic representation with Homer form to achieve efficient model computation. This invention reduces the algorithm development and complexity of five-axis machine tools and is suitable for the rapid deployment of general-purpose five-axis machine tool algorithms.

The invention belongs to the technical field of error detection, and particularly relates to a random-length parallel CRC efficient implementation method which comprises the following steps: S1, defining a CRC generator polynomial and representing the polynomial as an m * (m + 1) binary matrix G, decomposing the matrix G into sub-matrixes G0 of the first m columns and a sub-vector G1 of the last column, and initializing a CRC register state vector s with the length of m; and S2, according to the target parallel processing width N, constructing an m * (m + N) parallel transformation matrix T for processing N-bit data blocks at one time. According to the method, through a matrix T generation algorithm, the method can be suitable for any CRC standard defined by a matrix G and any parallel processing width, and the limitation of a traditional method on a specific polynomial or a fixed width is broken through.

An information processing apparatus of the present disclosure includes: a transforming unit that transforms a matrix included in a formulated model representing energy in a combinatorial optimization problem in such a manner as to decrease a rank of the matrix; a decomposing unit that decomposes the transformed matrix; and a solving unit that performs solution using the decomposed matrix.