65 results about "Matrix representation" patented technology

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 relates to the technical field of textile production scheduling and resource allocation planning, in particular to a textile equipment scheduling management and optimization system of a textile factory, which comprises a data perception and integration module, a scheduling optimization decision module, a plan execution and equipment control module and a closed-loop feedback and self-learning module. The data sensing module collects and fuses order data, equipment operation state data and production environment data in real time, and a unified real-time data view is generated through cleaning and alignment processing; the scheduling module runs a mixed integer programming dynamic model, and minimizes the comprehensive cost and synchronously optimizes the equipment utilization rate and energy consumption in combination with rolling horizon optimization under the condition of meeting the process constraints of order delivery time limit and process dependency matrix representation; the plan execution module analyzes the scheduling instruction into an equipment executable instruction, drives equipment operation and collects execution deviation; and the closed loop module triggers rescheduling when the deviation exceeds the limit or the order is plugged. The scheduling accuracy and adaptability are improved, the cost is reduced, and efficient and stable production is guaranteed.

The invention discloses a multi-modal knowledge graph representation method and device, and belongs to the field of computer vision representation. The method comprises the following steps: firstly, acquiring multi-modal embedding of a target entity, including a visual modal, a text modal and a structural modal, and calculating a Gram matrix formed by the multi-modal embedding; and calculating a determinant of the Gram matrix, and minimizing a geometric volume represented by the Gram matrix to coordinate semantic alignment among different modals and optimize multi-modal representation. A volume minimization constraint is introduced through a GraCon contrast learning framework, the geometric volume of multi-modal embedding is further optimized, and semantic consistency between different modals is ensured. Entity anchoring matching loss is introduced, and each entity is aligned by using a structural mode, so that the accuracy and consistency of entity representation are remarkably enhanced. And a Transform encoder is used to carry out reasoning and prediction of a knowledge graph completion task. The multi-modal representation quality can be improved, the entity representation is optimized, and the knowledge graph completion precision and robustness are improved.

The invention discloses a base station positioning method and system, and relates to the technical field of indoor positioning, and the method comprises the steps: obtaining target scene space structure information and signal propagation shielding characteristics which meet preset space constraints, building a scene shielding model, determining a candidate deployment region, and eliminating a high-attenuation shielding region. Acquiring a base station position Pi and a distance Lij between a label and the base station in the candidate area, and constructing a label position solving equation containing vector and matrix representation; and determining a base station layout avoidance scheme, deploying base stations, re-collecting data, correcting distance data, substituting the corrected distance data into the equation to obtain a label accurate position by combining the condition that the matrix determinant is close to zero and a scene shielding model, and obtaining a label positioning result. The system comprises a scene modeling module, a candidate area determination module, a data acquisition and equation construction module, a base station layout optimization module, a distance data correction module and a label position solving module, and solves the problems of unreasonable layout and large distance measurement error of the existing method.

The invention relates to a data commodity recommendation method and system based on deep learning, and the method comprises the steps: carrying out the word embedding coding of a user text and a commodity text of a data commodity through a first deep learning model, and obtaining a user text matrix representation and a commodity text matrix representation; respectively carrying out feature extraction and feature compression on the user text matrix representation and the commodity text matrix representation by utilizing a second deep learning model to obtain a user text vector representation and a commodity text vector representation; fusing the user text vector representation and the commodity text vector representation to obtain a fusion vector; and calculating a prediction score of the data commodity according to the fusion vector. According to the method, the data commodity recommendation effect can be improved, the recommended data commodity can meet the purchase demand of the user, and the user experience and the transaction efficiency are improved.

The invention relates to the technical field of multi-modal model reasoning, in particular to a visual language interpretable learning method and system assisted by neural logic reasoning. The method comprises the following steps: carrying out vector representation on an image-text pair; performing relation matrix representation on the first-order logic; performing logic combination and multi-hop reasoning based on the first-order logic and the representation of the image-text pair; based on logic combination and multi-hop reasoning, a logic reasoning micro attention network fused with a visual language model is constructed. According to the method, the LogicVLM model is designed, so that a microframework combining neural inductive learning and logical reasoning can be realized. And further learning first-order logic and logic combination from the input visual text semantic concept, constructing a path of a tree structure to execute multi-hop reasoning, and completing a complex visual language reasoning task. In an experiment, compared with a traditional visual language model, the model training speed and the reasoning speed of the method are basically kept unchanged, and various tasks are remarkably improved.

The invention discloses a coding and decoding method for LDPC codes under time and frequency selective channels in an NAVDAT system, and the method comprises the steps: determining a basis matrix and an expansion factor Z according to an information bit length and a target code length of an input bit sequence before coding, and the Z is a positive integer; obtaining a low density parity check (LDPC) matrix based on the basis matrix and the expansion factor Z; the method comprises the following steps: encoding an input bit sequence by using a low density parity check (LDPC) matrix under an NAVDAT system time and frequency selective channel, wherein the code rate is 3 / 4 or 1 / 2; wherein a basis matrix of the LDPC matrix comprises a sub-matrix A and a sub-matrix B, if the basis matrix is expressed as an m-row n-column matrix, the sub-matrix A is an m-row n-m-column matrix, the sub-matrix B is an m-row m-column matrix, and the sub-matrix B comprises a double-diagonal structure matrix B1, a single-diagonal structure matrix B2 and a column B3 with the weight of 4; and outputting the coded or decoded bit sequence. According to the invention, the coding requirements of information bit sequences with various lengths can be met.

The invention discloses a seismic source positioning method considering the influence of an abnormal value and a dynamic wave velocity, and the method comprises the steps: arranging sensors to detect sound waves emitted by a seismic source, and observing the time when the sound waves reach each sensor; subtracting the control equation established based on the reference sensor from the control equation established based on the measurement sensor, and carrying out linearization and matrix representation on a nonlinear equation line obtained by subtraction; the matrix is solved, an analytic solution enabling the sum of squares of the residual errors to be minimum is found, the obtained analytic solution is substituted into an overdetermined matrix equation to obtain a fitting model, and weight estimation is conducted according to the relation between the residual errors and weights; iteration is carried out between weight estimation and overdetermined matrix solving, after optimal weight estimation is obtained, overdetermined matrix solving is completed according to the optimal weight, and a final seismic source positioning result is obtained based on an analytical solution. According to the method, the technical problem that the influence of abnormal values and wave velocity measurement errors on the seismic source positioning precision in a complex engineering environment is large is solved, and the fault tolerance and the real-time performance of seismic source positioning are improved.

The invention discloses a community discovery method of a graph neural network based on multi-view information fusion, and belongs to the technical field of data mining. The method comprises the following steps: preprocessing data, performing PCA dimension reduction on an attribute matrix, and normalizing an adjacent matrix; inputting the correlation matrix combination into an automatic encoder and a graph attention automatic encoder to obtain a new matrix representation and fusing the new matrix representation into a final node representation matrix; then back-propagating the optimization model by using various loss functions; and finally, community division is realized by using a k-means clustering algorithm. According to the method, the structure and attribute information of the original data are fully utilized, multi-view information is effectively fused, the accuracy of community discovery is improved, and the method has great significance in the fields of research on citation networks, recommendation systems and the like.

Provided is a dynamic time warping (DTW)-based real-time diagnosis method for a commutation failure (CF). The method includes: designing a time series template S0 containing a change of a firing angle of a phase-controlled rectifier, and having a length n and a Euclidean distance L; using an n*n matrix M to represent a Euclidean distance between each point in the S0 and each point in the Si, and constricting the matrix M with an Itakura window; calculating values of the matrix M in the search range H; setting different weights for data at different locations, and searching a shortest path L* from point (x1, y1) to point (xn, yn) of the matrix; and setting a range of the distance L of the S0 as λ, determining that the CF fault occurs if the L* is greater than λL, or otherwise, determining that a system works normally.

The application provides a decoding method, an encoding method, a device, a chip and a computing device. The method comprises the following steps: a decoding device acquires a first matrix, and then decodes a first code word vector according to the first matrix to obtain a decoding result. The first matrix is a 0-1 matrix representing elements in a second matrix on a polynomial ring. The second matrix is a check matrix on a finite field 2q, and the number of rows is r and the number of columns is n. A first polynomial corresponding to the polynomial ring is a q-th irreducible polynomial on a binary field. The number of 1s in the first matrix is less than the number of 1s in the 0-1 matrix representation of the second matrix. The first code word vector comprises k information code elements and r check code elements, and n=k+r. Each code element comprises q data units, and each data unit comprises at least one bit. The scheme can effectively reduce the hardware implementation complexity and cost of RS encoding and decoding.

The invention relates to the technical field of kinetic analysis of mechanical structures, in particular to a structural modal parameter identification method based on joint low-rank tensor decomposition, which comprises the following steps: performing tensor modeling on expression X = phi Q of structural response under modal coordinates, performing segmentation operation on collected vibration response, and expressing the vibration response in a high-order tensor form; then, low-rank representation is carried out on each column of modal shape vectors and low-rank representation is carried out on each row of modal responses by utilizing the low-rank performance of the modal response vectors and the modal shape vectors; further, the combined low-rank tensor of the high-order tensor is decomposed and solved, and a matrix representation phi reflecting a modal shape and a matrix representation Q of modal response are obtained; and finally, extracting modal frequency and damping ratio information from the obtained Q. According to the structural modal parameter identification method, the dynamic parameters such as the structural modal frequency and the modal shape are effectively identified by using the low-rank characteristics of the modal response vector and the modal shape vector.

The application provides a power load anomaly detection method and device, and belongs to the technical field of power grid safety. The method comprises the following steps: standardizing power load data and processing the power load data in blocks. Then, a multi-head self-attention network is used to calculate an attention matrix representation between blocks and within blocks, and the attention matrix representation is respectively up-sampled. Next, a divergence loss function of the two is calculated, and an anomaly score of each point is calculated according to the divergence loss function. Finally, whether the power load data is abnormal is determined through a preset hyperparameter threshold. Through the method, the distance between normal and abnormal user features is maximized, the distance between features of the same type of users is minimized, an effective power data representation is learned, different user power consumption data features are actively compared, and abnormal power consumption behaviors are effectively identified.

Certain examples of the present disclosure relate to an apparatus (10, 120) comprising: means (11, 15) for sending, to one or more User Equipment, UE (110), demodulation reference signal, DMRS, matrix information (401) indicative of a row-orthogonal DMRS matrix (P); means (11, 15) for sending, to a UE (UEi) of the one or more UE, DMRS submatrix information (402) indicative of a submatrix (Pi) of the row-orthogonal DMRS matrix that has been assigned to the UE; means (11, 15) for receiving, from the UE, said at least one DMRS determined by the UE based, at least in part, on the DMRS matrix information and the DMRS submatrix information.

Methods, systems and apparatus for simulating physical systems. In one aspect, a method includes the actions of selecting a first set of basis functions for the simulation, wherein the first set of basis functions comprises an active and a virtual set of orbitals; defining a set of expansion operators for the simulation, wherein expansion operators in the set of expansion operators approximate fermionic excitations in an active space spanned by the active set of orbitals and a virtual space spanned by the virtual set of orbitals; performing multiple quantum computations to determine a matrix representation of a Hamiltonian characterizing the system in a second set of basis functions, computing, using the determined matrix representation of the Hamiltonian, eigenvalues and eigenvectors of the Hamiltonian; and determining properties of the physical system using the computed eigenvalues and eigenvectors.

The invention discloses a layered firepower planning method fusing a diffusion model and a chaos polynomial, and belongs to the field of spacecraft manufacturing and application. The target damage expectation evaluation method including energy field rasterization unified matrix representation, diffusion model energy field prediction and chaotic polynomial damage expectation calculation is established, and the efficiency and convenience of damage effect evaluation are improved. And counting energy field information contained in each grid based on a rasterization preprocessing method of the maximum range of the energy field. A hierarchical optimized firepower planning architecture is established, aiming point planning efficiency is improved, a top layer determines a distribution relation matrix of the aircraft and the target for firepower distribution, meanwhile, the incidental damage effect of the aircraft on the target is considered by utilizing a grouping strategy, and on the basis, a bottom layer aims at each group; and aiming point optimization is carried out by using a rapid and accurate target damage expectation evaluation method, an aiming point planning result is rapidly obtained, and the aircraft efficiency is exerted in high real-time performance according to the aiming point planning result.

The present invention discloses a network embedding method, device, electronic device, and computer program product based on a federated learning network. The method comprises: obtaining local network matrices of multiple data holders in a federated learning network, wherein the data holders are participants in the federated learning network, the federated learning network is pre-divided into multiple local learning networks, and the local network matrix is ​​a matrix representation of the local learning network; encrypting the local network matrix using a preset mask matrix to obtain a local encryption matrix; uploading the local encryption matrices corresponding to the multiple data holders to an embedding server, and receiving a global encryption matrix obtained by integrating the multiple local encryption matrices from the embedding server; and decrypting the global encryption matrix based on the preset mask matrix to obtain a global embedding matrix. The present invention solves the technical problem of poor data security in network embedding in a distributed environment in the prior art.

This invention relates to the field of drawing recognition technology, and discloses a method, apparatus, device, and readable storage medium for beam span recognition in vector graphics. The method includes: acquiring a target vector graphic and its corresponding matrix representation; identifying key beam span points from the matrix representation; determining the beam span type based on these key points; and mapping the positions of the key beam span points to the target vector graphic to obtain beam span elements corresponding to the beam span type in the target vector graphic. By implementing this invention, the recognition of beam span elements is not limited by the drawing method, ensuring the accuracy of beam span element recognition. This enables accurate identification of the positional information of beam elements, further guaranteeing accurate identification and attribute matching of beam elements.

The invention provides a security detection method and device for network traffic data, and the method comprises the steps: extracting a plurality of preset network features of each traffic data packet in a target time period, and obtaining a plurality of pieces of evolution data; dividing the plurality of evolution data of the target time period into to-be-detected data of a plurality of windows by using a sliding window; processing the to-be-detected data of each window to obtain a matrix representation of the to-be-detected data; determining an approximation ratio test value distribution based on the matrix representation of the data to be detected; determining a threshold value based on the approximate ratio test value distribution and a preset false alarm probability; whether the to-be-detected data is larger than a threshold value or not is judged, and if yes, it is determined that the to-be-detected data is abnormal data; and processing the abnormal data by utilizing hash function calculation and bit operation, and judging whether the abnormal data is attack data or not according to a processing result. According to the invention, the detection efficiency and detection precision of the network traffic data are improved.

The invention discloses a coding and decoding method for LDPC codes under a slight fading Gaussian channel in an NAVDAT system, and the method comprises the steps: determining a basis matrix and an expansion factor Z according to an information bit length and a target code length of an input bit sequence before coding, and the Z is a positive integer; obtaining a low density parity check (LDPC) matrix based on the basis matrix and the expansion factor Z; the method comprises the following steps: encoding an input bit sequence by using a low density parity check (LDPC) matrix under a slight fading Gaussian channel of an NAVDAT system, wherein the code rate is 3 / 4 or 1 / 2; wherein a basis matrix of the LDPC matrix comprises a sub-matrix A and a sub-matrix B, if the basis matrix is expressed as an m-row n-column matrix, the sub-matrix A is an m-row n-m-column matrix, the sub-matrix B is an m-row m-column matrix, and the sub-matrix B comprises a double-diagonal structure matrix B1, a single-diagonal structure matrix B2 and a column B3 with the weight of 4; and outputting the coded or decoded bit sequence. According to the invention, the coding requirements of information bit sequences with various lengths can be met.

The invention discloses a coding and decoding method for LDPC codes under multi-hop sky wave propagation in an NAVDAT system, and the method comprises the steps: determining a basis matrix and an expansion factor Z according to an information bit length and a target code length of an input bit sequence before coding, and the Z is a positive integer; obtaining a low density parity check (LDPC) matrix based on the basis matrix and the expansion factor Z; under NAVDAT system multi-hop sky wave propagation, an LDPC matrix is used to encode or decode an input bit sequence; wherein a basis matrix of the LDPC matrix comprises a sub-matrix A and a sub-matrix B, if the basis matrix is expressed as an m-row n-column matrix, the sub-matrix A is an m-row n-m-column matrix, the sub-matrix B is an m-row m-column matrix, and the sub-matrix B comprises a double-diagonal structure matrix B1, a single-diagonal structure matrix B2 and a column B3 with the weight of 4; and outputting the coded or decoded bit sequence. According to the invention, the coding requirements of information bit sequences with various lengths can be met.

The present invention provides a training method, device and electricity consumption information prediction method for an electricity consumption information prediction model, which relates to the fields of Internet technologies such as artificial intelligence and deep learning. The method comprises: inputting a sample quadruple consisting of a head entity object representing an electricity consumption area, a relationship object representing an electricity consumption unit value attribute, a tail entity object representing an electricity total value attribute and a time object representing timestamp information into a first neural network to obtain a sample embedding matrix representation of the sample quadruple; inputting the sample embedding matrix representation into a second neural network to obtain the sensitivity feature change rate of each of the electricity consumption area, the electricity consumption unit value attribute and the electricity total value attribute; inputting the sensitivity feature change rate into a third neural network to calculate a matching evaluation value prediction result of the predicted information relative to other real information; and calculating the training loss and adjusting the model parameters based on the sample embedding matrix representation, the matching evaluation value real label and the matching evaluation value prediction result.

The invention provides a CNN model parameter accelerated extraction method and system in a black box hard tag environment, and belongs to the technical field of artificial intelligence. Comprising the following steps: reconstructing a convolution kernel of a convolutional neural network into a matrix representation with a BTTB sparse structure based on structure information of the convolutional neural network; a double-point set is searched and determined through query and decision boundary geometrical shape analysis; and executing a clustering algorithm taking a convolution kernel as a center in the calculation unit to generate a double-point cluster, reconstructing a critical hyperplane by combining subspaces corresponding to each double point in the cluster, and extracting a normal vector of the critical hyperplane as unsigned estimation so as to determine the weight of a convolution layer. According to the method, through BTTB matrix reconstruction, clustering with the convolution kernel as the center and multi-stage optimization of covariance matrix acceleration, the memory overhead in the calculation process can be remarkably reduced, and the CNN model parameter extraction efficiency of a GPU processor is effectively improved.

The invention discloses a fully expressive sparse matrix representation with finite element data. Systems and techniques for compressing a dense matrix into a fully expressive sparse matrix representation with finite element data are disclosed. The techniques include generating a sparse matrix with corresponding metadata based on the dense matrix. Generating a sparse matrix with respective metadata includes: identifying a first number (M) of elements to be compressed, a second number (N) of elements to be retained, and a third number (B) of bits used by each metadata value; determining a metadata value of each of the N elements of the dense matrix; packing the first metadata value having more than B bits into a second metadata value having B bits; and generating a sparse matrix containing the N elements of the dense matrix. The techniques include storing a sparse matrix and corresponding metadata, wherein the corresponding metadata includes a second metadata value.

The invention discloses arithmetical operations using expressive sparse matrix representations with finite element data, and specifically discloses systems and techniques for performing matrix multiplication operations on expressive sparse matrix representations with finite element data. The techniques include receiving a sparse matrix, metadata corresponding to the sparse matrix, and a matrix operand. The sparse matrix contains a first number (N) of elements to be retained in a dense matrix comprising at least a second number (M) of elements. The metadata corresponding to the sparse matrix is based on a third number (P) of positions and a format determined during compression of the dense matrix. The techniques include selecting, by one or more selection circuits, a subset of elements of a matrix operand based on metadata corresponding to a sparse matrix, and performing one or more matrix multiplication operations on the sparse matrix and the subset of elements of the matrix operand.

In order to carry out an improved image correction, it is provided that a number of NM multiplication influence factors and a number of NC convolution influence factors are taken into account in the image recording (1), said multiplication influence factors describing multiplication factors between image points (p) in the interference-reduced image (U) and the same image points (p) in the interference-affected image (F), each multiplicative influence factor is described in a vector-matrix representation as a single multiplicative matrix (Mi) in the form of a diagonal matrix having matrix elements for each pixel (p) on a diagonal line, and the multiplicative matrix M is determined as a matrix product of the single multiplicative matrixes (Mi) from a number NM of the single multiplicative matrixes (Mi), and wherein the convolution influence factors describe the influence of a plurality of image points (p) in the interference reduction image (U) on an image point (p) in the interfered image (F), and each convolution influence factor is described in the form of a single convolution matrix (Ci), each row of individual convolution matrices (Ci) describes how an image point (p) in the interfered image (F) is affected by a respective image point (p) in the interference-reducing image (U), and a convolution matrix C is determined as a matrix product of the individual convolution matrices (Ci) from a number of NC individual convolution matrices (Ci), and wherein the matrix product of the multiplication matrix M and the convolution matrix C is used as a correction operator (A) for image correction, i.e. A = MC.

Disclosed are systems and techniques for decompressing an expressive sparse matrix representation with limited metadata. The techniques include receiving a sparse matrix and metadata corresponding to the sparse matrix. The sparse matrix is a compressed representation of a dense matrix. The sparse matrix contains a first number (N) of elements to retain from the dense matrix which comprises at least a second number (M) of elements. The metadata corresponding to the sparse matrix is based on a third number (P) of positions and a format determined during compression of the dense matrix. The techniques include generating an uncompressed matrix based on the sparse matrix and the metadata corresponding to the sparse matrix.

The application discloses a kind of array antenna wave direction, time delay and carrier frequency offset joint estimation method, comprising: the signal receiving model of array element is established, based on signal receiving model, the matrix representation form of each array element received signal is constructed and is handled, and the signal matrix of each array element is constructed;Signal group is divided according to subcarrier, and the signal matrix of each array element is stacked, and the stacked matrix is constructed;The stacked matrix is written into the form of tensor and is parallel factor decomposition, and the matrix containing DOA, time delay and carrier frequency offset is obtained respectively;From the matrix after parallel factor decomposition, DOA and time delay are estimated by spectral peak search and least square method according to the corresponding relationship respectively, the pairing of DOA and time delay is realized;And carrier frequency offset is estimated using the matrix after decomposition.The method of the application has high estimation accuracy for each parameter, and better robustness.

The invention discloses a cross-format receipt semantic content extraction method based on layout semantic nodes, and belongs to the technical field of computer application, and the method comprises the following steps: S1, constructing a keyword dictionary; s2, defining a semantic content extraction template, and representing the semantic content extraction template by using a local relation direction matrix; the documents are recognized through OCR, and a global relation direction matrix is constructed according to the keyword dictionary; s3, global-local matrix matching is carried out to position a target area; and S4, semantic extraction position determination, text block screening and structured output. According to the cross-format receipt semantic content extraction method based on the layout semantic nodes, fault tolerance is improved, low-cost and fast deployment is achieved without large-scale data labeling, auditing performance is guaranteed through explicit modeling, and the mismatching rate is reduced; the whole text processing process is covered through a parameterization strategy, complex scenes such as cross-column and broken-line are efficiently dealt with, and the blank of weak adaptation capability of complex layout is filled up.

The present invention proposes a quantum circuit optimization method, device and equipment for a subcomponent of the Zu Chongzhi algorithm. The method comprises: encoding a communication transmission signal into quantum bits and inputting the quantum bits into the Zu Chongzhi algorithm structure for quantum implementation; representing the first linear subcomponent in the Zu Chongzhi algorithm structure as a matrix on a binary domain; representing the matrix on the binary domain as a block matrix, and deriving a low-dimensional matrix on the binary domain from the matrix. M , using a search algorithm designed based on the principle of matrix decomposition, search M According to the quantum gate parallel rule, M The quantum implementation of re-optimized multi-layer partitioning is obtained M The quantum implementation of the algorithm optimizes the hierarchical operation and optimizes the quantum implementation of the first linear subcomponent. The method of the present invention optimizes the quantum circuit of the subcomponent of the Zu Chongzhi algorithm, thereby saving the resource overhead of the quantum implementation of the Zu Chongzhi algorithm.