16 results about "Sparse matrix factorization" patented technology

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.

The invention relates to a storage calculation device supporting sparse matrix calculation, which comprises a storage core particle, a logic core particle, a multiply-accumulate circuit, a characteristic value distribution circuit and a sparse matrix decomposition circuit, and is characterized in that the storage core particle is used for storing a sparse weight matrix, and the logic core particle is used for data interaction, caching and control; the multiply-accumulate circuit executes multiply-accumulate operation of characteristic values and weights, the sparse matrix decomposition circuit screens out position codes of front # MAC non-zero weights in each operation period and outputs control signals, and the characteristic value distribution circuit dynamically maps input characteristic values to corresponding non-zero weights according to the control signals, so that calculation of an arbitrarily distributed sparse matrix is realized. The method has the advantages that multiply-accumulate operation is performed on the non-zero weight, the zero weight is skipped, power consumption is reduced, calculation efficiency is improved, and efficient processing of the sparse matrix is achieved through a storage and calculation integrated framework.

The invention discloses a main and branch metering user multi-grid-connected point photovoltaic precise metering method, which comprises the following steps: step 1, obtaining an impedance value at a photovoltaic power generation metering point, and obtaining an H-class impedance sample set; 2, obtaining a sparse impedance value by using a K-SVD sparse decomposition algorithm, and constructing a sparse matrix; step 3, updating the discrimination dictionary based on the sparse matrix; 4, decomposing the sparse matrix into a trend term and a periodic term by using an Autoformer model encoder; 5, decomposing the impedance value in real time by using an Autoformer model to obtain real-time mains supply impedance, photovoltaic impedance, sub-metering impedance and other part impedance under different frequencies; and step 6, obtaining the impedance value of each part under the power frequency, and realizing the multi-grid-connection-point distributed photovoltaic accurate metering of the main and branch metering users. According to the invention, no extra hardware is added, and the physical quantity is obtained by using the carrier signal of the electric energy meter to realize accurate metering.

The invention relates to the technical field of computer hardware quality detection, and discloses a computer mainboard quality detection method and system based on artificial intelligence, and the method comprises the following steps: collecting multi-modal data of a to-be-detected computer mainboard; constructing a multi-modal time sequence tensor according to the timestamp; inputting the time sequence tensor into a time sequence neural network, and extracting a dynamic feature map reflecting defect evolution; performing low-rank sparse matrix decomposition to obtain an abnormal feature map; calculating the defect saliency of the abnormal feature map; when the defect enhancement trend is judged, feedback detection is executed; the system comprises an image acquisition module, a thermal imaging acquisition module, an electromagnetic induction module, a data preprocessing module, a time sequence analysis module, an anomaly detection module, a trend judgment module and a feedback control module. According to the method, the feature graph decoupling technology based on low-rank sparse matrix decomposition is introduced, and a more accurate abnormal feature graph can be extracted from the background structure and the potential defect area of the mainboard.

The application discloses a course teaching resource recommendation method, comprising the following steps: establishing a course teaching resource recommendation dataset, learning text and image features of course content and course teaching resources in a training set, and calculating a similarity matrix between the text and image features; based on the similarity matrix and existing scoring samples, a low-rank sparse matrix decomposition recommendation model of the course teaching resources is constructed, and a scoring matrix is decomposed into a low-rank latent matrix related to course content and a sparse latent matrix related to course teaching resources; the low-rank matrix and the sparse matrix are initialized by using random numbers, the low-rank matrix and the sparse matrix are iteratively optimized by using a gradient descent method, and the optimization is updated until a target function based on the similarity matrix, the low-rank matrix and the sparse matrix converges; course content features in test data are extracted, corresponding course teaching resources are found from approximate estimation of the scoring matrix, and a recommendation list is returned according to scoring item sorting.

The invention relates to an SLAM back-end optimization method based on sparse matrix factorization and a hardware accelerator architecture, and the method comprises the steps: constructing a Hessian matrix through employing a binary sparse coding mechanism according to the actual observation coordinates of feature points on an image plane and the predicted coordinates of road sign points projected to a camera coordinate system; performing Schel elimination processing on the Hessian matrix based on binary sparse coding to obtain a linear equation set only containing a camera pose variable; and carrying out extended Kalman filtering on a linear equation set only containing a camera pose variable to realize incremental updating of a state pose and a road sign point coordinate. According to the method, the overall delay of SLAM rear-end optimization can be remarkably reduced, and the real-time requirement in a high-speed moving scene is met.

The invention provides a user portrait construction method under an AI platform, and belongs to the technical field of user portray.The method comprises the steps that microscopic interaction data are obtained through a user behavior collector, and a behavior state vector set is constructed according to a time sequence; a probability model and an interest energy function are established based on a Maxwell-Boltzmann distribution principle to automatically identify and filter noise data, and sparse matrix decomposition is adopted to project filtered clean data to a low-dimensional semantic feature space. A semantic enhancement model is utilized to perform clustering analysis on high-quality behavior data to identify a real user interest mode, a user individual graph structure based on semantic association is constructed through a knowledge graph technology to accurately express user preferences, and MCP protocol packaging is adopted to realize fusion processing of clean microscopic behaviors and accurate macroscopic features. The technical problem that the accuracy of the formed user portrait is not enough due to the fact that the user microcosmic behavior data is easily interfered by noise is solved.

The present application relates to the technical field of signal processing, in particular to a kind of sparse low-rank decomposition DOA estimation method based on virtual array interpolation, establish receiving signal model;The sampling covariance matrix of receiving signal model is calculated, and vectorization operation is carried out to sampling covariance matrix, since there is hole for a non-uniform virtual array in virtual array, virtual sensor is inserted at the hole position of virtual array, form virtual uniform array, reconstruct virtual uniform array receiving signal model as Toeplitz matrix, and carry out low-rank and sparse matrix decomposition into signal covariance matrix and noise covariance matrix;Signal covariance matrix is reconstructed using low-rank matrix recovery theory, and the expected signal covariance matrix is recovered, it is carried out eigenvalue decomposition using MUSIC algorithm, and the target DOA is estimated.This application makes full use of all virtual array element information, avoids the inaccuracy of DOA estimation caused by information loss.

The invention discloses an image data uncertainty quantification method based on low-rank sparse matrix factorization, which comprises the following steps: an image data preprocessing stage: collecting and matrix image data: flattening a plurality of polluted images into column vectors and arranging the column vectors in sequence to form an observation matrix; the number of rows of the observation matrix corresponds to the number of pixel points of the polluted images, the number of columns of the observation matrix corresponds to the number of the polluted images, missing marking, normalization and observation probability estimation are carried out on the observation matrix, and then an observation index set is randomly divided into a training set and a calibration set in proportion; in the training stage, a non-distributed robust principal component analysis method based on a conformal prediction framework is used for processing the training set, low-rank estimation, residual standard deviation and a weighted threshold are obtained through training, and a low-rank structure uncertainty recovery model is obtained; a test stage: performing low-rank recovery and uncertainty quantification on new polluted image data by using the low-rank structure uncertainty recovery model; the method is suitable for data recovery and uncertainty quantification in a complex environment.

The invention discloses a deep expansion network-based remote sensing video satellite motion small target detection method, which is applied to the field of target detection and comprises the following steps of: decomposing a low-rank sparse matrix decomposition task into a background solving task, a foreground solving task and a Lagrange multiplier updating task based on an alternating direction multiplier method; constructing a background recovery layer based on a background solving task, constructing a foreground reconstruction layer based on a foreground solving task, and constructing a Lagrange multiplier updating layer based on a Lagrange multiplier updating task; constructing a deep expansion network based on the background recovery layer, the foreground reconstruction layer and the Lagrange multiplier update layer, and training the deep expansion network based on remote sensing video satellite video data; and after the training is completed, inputting to-be-detected video data into the deep expansion network to obtain an output moving small target detection result. Priori knowledge is brought into the network to guide network learning through the deep expansion network, and the method has high operation efficiency and good generalization ability while maintaining interpretability and low parameter sensitivity.

The present application belongs to the technical field of sparse matrix decomposition, and relates to a block method of performance-aware block sparse LU decomposition for a GPU platform, comprising the following steps: S1, constructing a performance function of a GPU computing kernel based on statistical data; S2, based on the performance function, designing a decision method for deciding whether to block a sparse matrix at a specified position; and S3, decomposing the overall block problem of a large-scale sparse matrix into multiple sub-problems decided in sequence, solving each sub-problem one by one based on a greedy heuristic strategy and using the decision method, and obtaining a final specific block scheme. Compared with existing methods, the method of the present application is more excellent in performance in the numerical decomposition stage, and is not only suitable for a heterogeneous platform of multi-core CPU and GPU, but also suitable for a pure CPU architecture.

This invention proposes a clutter suppression method suitable for pseudo-randomized ground-penetrating radar (GPR), belonging to the field of GPR data processing. It includes: converting A-SCAN data into a two-dimensional square matrix; using the equivalent matrix gamma norm and equivalent minimum-maximum concavity penalty as non-convex substitutions for rank and sparsity; low-rank sparse matrix decomposition; and A-SCAN stitching imaging. This method addresses the problem of inaccurate rank estimation caused by directly using large-scale B-SCAN data in existing clutter suppression methods. It uses a single A-SCAN as the smallest processing unit, arranging them sequentially into a two-dimensional square matrix to achieve accurate rank estimation. By using the equivalent matrix gamma norm and equivalent minimum-maximum concavity penalty as non-convex substitutions for rank and sparsity, it achieves a more accurate estimation of rank and sparsity compared to traditional convex substitution methods. Finally, clutter suppression results are obtained through low-rank sparse matrix decomposition, achieving clutter suppression for pseudo-randomized ground-penetrating radar.

Embodiments of the present application provide a service matter recommendation method and device, electronic equipment and storage medium; the service matter recommendation method comprises obtaining public information and user information; wherein, the public information comprises public behavior and public service matters, and the user information comprises user behavior and user service matters; a public co-occurrence matrix is constructed based on the public behavior and the public service matters; a user co-occurrence matrix is constructed based on the user behavior and the user service matters; the user co-occurrence matrix is sparsified by using the public co-occurrence matrix, and a co-occurrence sparse matrix is generated; the co-occurrence sparse matrix is decomposed to determine a recommended service matter. Through the embodiments of the present application, the accuracy of pushing and recommending service matters to users can be improved.

The present disclosure relates to a polarization adaptive scattering suppression based underwater structured light reconstruction method and system, comprising: acquiring a phase shift fringe image sequence under four polarization angles; respectively processing each phase shift fringe image sequence through time domain integral averaging to obtain average light intensity images of each polarization state; calculating average polarization degree and average polarization angle images according to the average light intensity images, and performing low-rank sparse matrix decomposition to obtain a low-rank matrix; calculating the backscattering light intensity distribution based on the low-rank matrix and the average light intensity images, estimating the infinite background light intensity by using an adaptive iterative truncated mean algorithm, and then calculating the transmission coefficient of the scattering medium; correcting the phase shift fringe image sequence frame by frame according to the backscattering light intensity distribution and the transmission coefficient, obtaining the absolute phase through phase unwrapping, and reconstructing the three-dimensional point cloud of the underwater object to be measured. The present disclosure can effectively suppress scattering and reflection interference in a strong scattering, high turbidity and high reflection underwater environment, and improve the robustness and accuracy of three-dimensional reconstruction.

The application discloses an interactive point cloud data editing method on a mobile terminal and relates to the technical field of point cloud data processing, which comprises the following steps: collecting point cloud data in a transformer substation by using a handheld laser radar, performing preliminary processing on the point cloud data by using a mobile terminal, and segmenting the point cloud data by using a fragmentation algorithm; transmitting the fragmented point cloud data to an edge computing device through a high-speed wireless network, filtering noise of the point cloud data on the edge computing device by using an adaptive learning algorithm, and dynamically adjusting filtering parameters; extracting geometric features of the point cloud data filtered by noise on the edge computing device, extracting geometric feature points and performing compression processing; receiving the compressed geometric feature point data returned from the edge computing device by the mobile terminal, decompressing and reconstructing the point cloud data by using an inverse sparse matrix decomposition algorithm, and enabling a user to perform interactive editing operation by using the mobile terminal; and adjusting a display perspective of a point cloud model by the user through a gesture operation.

The invention discloses an electromagnetic transient simulation matrix solving method based on an NUMA (Non Uniform Memory Access) architecture and Shuercomplement, which relates to the technical field of electromagnetic transient simulation, and comprises the following steps: S1, carrying out row and column rearrangement on a sparse dispersion nano matrix of electromagnetic transient simulation through permutation transformation, so that the sparse dispersion nano matrix is in a block diagonal edge adding form, a global solving problem is decoupled into a plurality of independent subdomain elimination problems and a low-latitude Schel complement matrix interface problem; and S2, establishing a mapping model of the computing sub-domain and the CPU physical architecture. According to the electromagnetic transient simulation matrix solving method based on the NUMA architecture and the Shuer complement, through an innovative matrix decomposition and hardware resource optimization strategy, the calculation efficiency of electromagnetic transient simulation is remarkably improved, the weak coupling characteristic of a power network is ingeniously utilized, and the calculation efficiency of the electromagnetic transient simulation is improved. A large-scale sparse matrix is decomposed into a plurality of independent low-dimensional sub-domain matrixes and a low-dimensional Schel complement interface matrix, and parallel processing of calculation tasks is achieved.