13 results about "Coordinate descent method" patented technology

This invention discloses a latency optimization method for end-edge collaborative inference systems based on fluid antenna assistance (FA). The method first establishes an FA-assisted system model and constructs DNN partitioning collaboration, IFD transmission, and inference accuracy models based on the system model to determine the total inference latency, including local inference, IFD wireless transmission, and edge inference delays. Then, a joint optimization problem is constructed with the objective of minimizing the total inference latency, jointly optimizing parameters such as DNN partitioning points, transmit power of each device, and computational resource allocation. The joint optimization problem must satisfy inference accuracy and energy consumption constraints. Finally, the problem is decomposed into four sub-problems, and the block coordinate descent method is used to alternately optimize until convergence, completing the latency optimization of the end-edge collaborative inference system. This invention is the first to integrate FA and collaborative inference frameworks, achieving multi-dimensional parameter joint optimization, effectively reducing the total inference latency, and demonstrating significant performance advantages in scenarios with high accuracy thresholds and adverse channel conditions.

The invention discloses a current transformer error online monitoring method and system based on elastic network regression, and belongs to the technical field of current transformer error monitoring, and the method comprises the steps: collecting m groups of secondary fundamental current phasor values of n current transformers, and constructing an m * n-dimensional secondary current data matrix; establishing an n-dimensional overdetermined solution regression equation set based on the secondary current data matrix, the to-be-solved mutual inductor error parameter matrix and the residual vector; constructing an elastic network hybrid regularization loss function; solving the overdetermined regression equation set by adopting a coordinate descent method to obtain a mutual inductor error parameter matrix; and calculating the ratio difference and the angular difference of each mutual inductor according to the mutual inductor error parameter matrix. The technical problems that the error identification precision of an existing on-line monitoring device of the mutual inductor is not high enough, and multi-mutual inductor monitoring cannot be completed are solved.

The invention relates to the technical field of seismic data processing, in particular to a Marchenko imaging-oriented dual-norm constraint near-offset missing seismic data reconstruction method and device. The method comprises the following steps: obtaining original data, and carrying out Radon reconstruction on the original data to obtain data after Radon reconstruction; calculating a residual error between the Radon domain reconstruction data and the original data to obtain empty channel data; constructing a sparse reconstruction framework based on an lq1-lq2 norm to obtain a final objective function; and by utilizing a block coordinate descent method, substituting the data reconstructed based on the Radon and the empty channel data into the final objective function for solving so as to obtain a final reconstruction result. According to the method, a conventional single-norm sparse reconstruction method based on an l1 norm is improved into a double-norm constraint framework based on an lq1-lq2 norm, Radon reconstruction and Shearlet domain reconstruction are combined, compared with a traditional reconstruction method, the improved method has the advantages that the reconstruction effect of near large offset missing is obviously improved, and the reconstruction efficiency is greatly improved. And meanwhile, the requirement on the sampling rate of the input data is also reduced.

The present invention discloses a method for solving the inverse mapping of the Kriging proxy model of motor performance based on the coordinate descent method. First, different motor size parameter combinations are sampled, and the samples are used to construct a Kriging proxy model with multiple motor size parameters as input variables and a single motor performance as the output response. #imgabs0# Then, given the motor performance y c , #imgabs1# is used as the minimization objective function, and the range of motor size parameters in the samples from the Kriging proxy model training source is used as the feasible domain to construct the optimization model. Finally, the objective function is repeatedly minimized along the coordinate axis direction corresponding to the single component of the input variable based on the coordinate descent method, so that the objective function finally converges to 0, and the given motor performance y is obtained. c Corresponding motor size parameter combination #imgabs2#The method proposed in this invention can solve the inverse mapping of the Kriging proxy model of motor performance, and has good practicality and economy.

The present invention belongs to the field of wireless communication technology, and specifically discloses a method, device and medium for RSMA-assisted MEC system secure offloading; the method comprises: establishing an RSMA-assisted secure MEC system model containing an internal eavesdropper; establishing a target optimization model with the goal of minimizing the maximum delay of the secure MEC system model and taking the common rate allocation vector, precoding matrix and offloading ratio in the secure MEC system model as optimization variables; obtaining a sub-problem of jointly optimizing the precoding matrix and the common rate allocation vector by fixing the offloading ratio, and approximating the sub-problem of the first problem to a convex optimization problem by using a continuous convex approximation algorithm; obtaining a sub-problem of optimizing the offloading ratio by fixing the precoding matrix and the common rate allocation vector, and the sub-problem of the second problem being a convex optimization problem; and alternately iterating the approximated sub-problem of the first problem and the sub-problem of the second problem by using a block coordinate descent method until the iterative error of the iterative objective function is less than a threshold value, thereby obtaining the optimal delay of the target optimization model.

The application provides a multi-view face recognition system (MvDRHSFS) based on sparse feature selection, which can select the view with the most information and the feature from multiple views of data through a projection matrix to learn a low-dimensional subspace, thereby solving the dimension disaster problem of high-dimensional face images. First, the F norm is used to constrain the projection matrix to select the view with the most information. Second, the l21 norm is used to constrain the projection matrix to select the feature with the most information. The projection matrix and the recovery matrix are optimized by the coordinate descent method to obtain the optimal solution, and the low-dimensional space representation of the test face is obtained through the projection matrix. The application makes up for the deficiency of the existing multi-view dimension reduction method which can only select a single view or feature, and can select the view with the most information and the feature at the same time, thereby better guiding the learning of the low-dimensional subspace of the face and further improving the performance of the face recognition system.

This invention discloses a motor design method based on a Kriging surrogate model. Specifically, it involves: collecting M combinations of motor size parameters as input sample data S, and using the corresponding motor performance as output sample data Y; forming a sample set from S and Y; training a Kriging surrogate model based on this sample set; constructing an optimization model with minimization as the objective function and the range of motor size parameter variations in the samples from which the Kriging surrogate model is trained as the feasible region; and solving the optimization model to obtain the motor parameters corresponding to the given motor performance. The method proposed in this invention improves the speed of solving the inverse mapping of the multi-performance Kriging surrogate model of a motor based on the coordinate descent method, and has good practicality and economy.

This invention discloses a multi-dimensional resource management method for UAV-assisted edge computing networks, primarily addressing the problem that existing UAVs equipped with edge computing servers cannot provide stable services to a large number of users. The implementation scheme is as follows: 1) Initialize network parameters; 2) Set constraints based on network characteristics; 3) Establish a minimum energy consumption problem P based on the initialized network parameters and constraints; 4) Determine the correlation matrix C, offloading matrix U, computational resource matrix, and UAV position matrix Z in problem P using the block coordinate descent method; 5) Perform multi-dimensional resource management based on the obtained C, U, F, and Z values. This invention can optimize offloading decisions and deployment schemes based on user location and computational task information, reducing system energy consumption and solving the problem of user service obstruction, while meeting the constraints of tolerable latency of computational tasks and the upper limit of UAV computational resources. It can be used for emergency communication and computing services in UAV-assisted edge computing networks under conditions of limited UAV energy consumption.

The invention provides a multi-model reasoning service-oriented cross-layer collaborative scheduling method, which comprises the following steps of: firstly, determining the number of sub-graphs and an initial division point, then iteratively optimizing sub-graph segmentation by using a coordinate descent method, and selecting an optimal scheme in combination with a historical request weight and an actual time delay. Then, the sub-graph is divided into sub-graph blocks in a recursive mode, parallel scheduling schemes of different batch processing sizes are generated, and online candidate strategies are screened out; and meanwhile, new establishment, stretching and splitting operations are configured to realize dynamic batch processing, so that an optimal parallel strategy is scheduled and selected on line based on the service quality. The invention further provides a multi-model reasoning service-oriented cross-layer collaborative scheduling system, a storage medium and computer equipment. Therefore, the online model service performance can be improved, the computing resource utilization rate can be improved, online scheduling and offline optimization cooperative scheduling can be realized, and the service performance can be greatly improved.

The invention discloses a network anomaly detection method based on online tuck decomposition of meta learning enhancement. The method comprises the following steps: obtaining a multi-dimensional network traffic tensor at a t moment, carrying out tuck decomposition to obtain a factor matrix and a core tensor at the t moment, and carrying out dimension expansion; updating the incremental data at the t + 1 moment to obtain a factor matrix at the t + 1 moment; performing singular value decomposition on the multi-dimensional network traffic tensor at the t + 1 moment; counting the accumulated singular value and the minimum singular value number required for reaching 80% of the tensor norm of the moment t + 1 as the low-rank feature representation dimension of the moment t + 1; dynamically determining the rank of each dimension, and iteratively optimizing the core tensor until convergence; and constructing an anomaly detection framework, decoupling online network traffic anomaly detection into an alternate optimization tensor decomposition sub-problem and an anomaly detection sub-problem by adopting a block coordinate descent method, and performing network anomaly point identification according to the updated factor matrix and the core tensor in combination with a sparse residual screening mechanism. According to the invention, the abnormal flow can be accurately and rapidly detected.

The present invention discloses a method, system, medium and equipment for low-rank constrained reconstruction of five-dimensional seismic data, which comprises a data tensor, a sampling tensor, weight parameters of optimization terms and regularization terms, and the TR rank R1, ..., R1 of the matrix decomposed by the mode-{n, l} method along the direction. N Input seismic data to reconstruct the model and initialize the parameter Y 0 , #imgabs0# as the initial value; expand the low-rank reconstruction data tensor of the current step, cross-optimize the variables using the sampling block coordinate descent method during iteration, and use random sampling to improve computational efficiency. The low-rank reconstruction data is decomposed by mode-{n,l}#imgabs1#, and the optimal decomposition matrix #imgabs2# is calculated. The training sample #imgabs3# is reconstructed using the tensor to obtain the next step #imgabs4#. Repeat these steps until the predetermined stopping criterion is met, obtaining the low-rank reconstruction result of the real data. All frequency components are combined and restored to the original data format, achieving five-dimensional seismic data reconstruction. This method has good reconstruction effect, reduces computing resource consumption, and has broad industrial application prospects.

The invention discloses an underwater self-interference suppression method, which is based on a fast bisection coordinate descent recursive least squares algorithm (FDCD_RLS) with a variable forgetting factor. The method comprises the following steps: initializing algorithm parameters, including a forgetting factor, a step size, a number of iterations, an initial error, and an initial value of an autocorrelation matrix; calculating an instantaneous error value at a current moment by updating an input signal vector; calculating the forgetting factor at a current moment according to the instantaneous error value, wherein an inverse tangent function relationship exists between the value of the forgetting factor and the square of the instantaneous error; updating a ring index sequence, and updating the autocorrelation matrix and a pseudo error vector through the ring index sequence; updating a weight increment and a pseudo error vector through a bisection coordinate descent method; updating an estimated self-interference channel parameter at a current moment according to the weight increment; and repeating the above steps until convergence, thereby finally obtaining an estimated value of the underwater self-interference channel.

The present invention discloses a network anomaly detection method based on online Tucker decomposition enhanced by meta-learning. The method comprises the following steps: obtaining a multidimensional network traffic tensor at time t, performing Tucker decomposition to obtain a factor matrix and a core tensor at time t, and then dimensionalizing the data; updating the factor matrix at time t+1 using incremental data at time t+1; performing singular value decomposition on the multidimensional network traffic tensor at time t+1; calculating the cumulative singular values ​​and the minimum number of singular values ​​required to reach 80% of the tensor norm at time t+1 as the low-rank feature representation dimension at time t+1; dynamically determining the rank of each dimension, iteratively optimizing the core tensor until convergence; constructing an anomaly detection framework, using the block coordinate descent method to decouple online network traffic anomaly detection into alternating optimization of the tensor decomposition subproblem and the anomaly detection subproblem, and combining a sparse residual screening mechanism to identify network anomalies based on the updated factor matrix and core tensor. The present invention can accurately and quickly detect abnormal traffic.