36 results about "Matrix norm" patented technology

The invention discloses a ubiquitous power Internet of Things perception data missing restoration method based on matrix filling. The ubiquitous power Internet of Things perception data missing restoration method comprises the following steps: reconstructing measured one-dimensional time series data into a matrix form through slice transformation; obtaining a low-rank intensity index used for verifying data recovery feasibility; considering the structural characteristics of different components of the measurement data, establishing an optimization model for recovering missing data based on a low-rank matrix filling theory, constraining various noises through a matrix norm, and eliminating the noise influence; and obtaining an iterative calculation formula for quickly solving the model by improving an alternating direction multiplier method, and realizing recovery of missing measurement data. According to the ubiquitous power Internet of Things perception data missing restoration method, under the condition that part of the measurement data is lost and various forms of noise such as Gaussian noise and peak abnormal value are mixed, the original complete measurement data is recoveredbased on the low-rank matrix filling theory, and then complementation of the missing data is achieved.

The method includes following steps: after decoding, acquires norm NDCT of DCT coefficient matrix of current macro block / sub block; calculates the transformed norm Nq of quantify matrix set by the current macro block / sub block; calculates relevant quantify step size of DCT coefficient matrix being relative to transformation of quantify matrix DCT, a=Nq / NDCT; compares a with experience-threshold k; if a<k, then ends the process. If a>k, then makes smoothing process for current macro block / sub block and their neighboring pixels. The unit used in video encode and decode system consists of DCT coefficient matrix norm calculating device, quantify matrix DCT transformation calculation device, relevant quantify step size calculation device, comparator and smooth-processing device.

The invention relates to a regularized multi-view clustering machine learning method derived from matrix norm. The specific steps of the method include: 1) obtaining clustering task and target data samples; 2) deriving a regularization term based on that obtain matrix norm of the clustering task; 3) deriving a regularization term based on that matrix norm, and establishing a regularization multi-view clustering optimization objective function; 4) the regularized multi-view clustering optimization objective function is solved in a cyclic way to realize clustering. Compared with the prior art, the invention reduces the redundancy of selected cores and increases the diversity of selected cores by measuring the correlation between each pair of cores, and has the advantages of improving the clustering effect and the like.

The invention discloses a complex mode evaluation method of a finite element model with a damping structure. The method includes the steps: firstly, calculating a complex-frequency correlation coefficient according to relative errors of test complex frequencies and simulation complex frequencies, calculating a complex mode amplitude correlation coefficient of test and simulation by the aid of model assurance criterions, and calculating a complex mode phase position correlation coefficient by the aid of defined matrix norms; secondly, performing fuzzification on correlation coefficients of the complex frequencies, complex mode amplitudes and complex mode phase positions by the aid of a trapezoidal membership function to obtain fuzzy relation matrixes of three influencing factors, and performing mathematic operation on the fuzzy relation matrixes and all-order modal weight sets to obtain evaluation results of the influencing factors; finally, performing mathematic operation on the evaluation results and weight sets of the three influencing factors to obtain a comprehensive evaluation value. According to the method, influence of damping on structural response is considered, the method is applicable to correction of models with damping structures, error position identification and dynamic response accuracy of prediction models, calculation methods of the models are normative to implement, and implementation is facilitated by the aid of computer programming.

A convex regularized loss function is minimized respective to a prediction tensor of order K to generate an optimized prediction tensor of order K where K>2. The convex regularized loss function comprises a linear combination of (i) a loss function comparing the prediction tensor and an observation tensor of order K representing a set of observations and (ii) a regularization parameter including a K-th order matrix norm decomposition of the tensor trace norm of the prediction tensor. In some such embodiments, the observation tensor of order K represents a set of social network observations and includes at least dimensions corresponding to (1) users, (2) items, and (3) tags. The optimized prediction tensor of order K is suitably used to perform inference operations.

The invention provides a stratospheric large-scale MIMO user grouping and precoding method and system, and the method comprises the steps: an optimization problem design step: designing an optimization problem according to an outer-layer precoding matrix and an inner-layer precoding matrix on the basis that it is assumed that multiple users of a stratospheric large-scale MIMO communication systemare grouped; Optimization problem simplification step: carrying out optimization problem simplification; According to the method, a matrix norm inequality and a zero space criterion are adopted, an obtained optimization problem is simplified, the power of a signal is explored to be mainly concentrated on a statistical intrinsic mode StacticalEgigades of a channel, and the statistical intrinsic mode StacalEgigades are obtained. In the invention, the CSI of the effective channel for designing the inner-layer precoding matrix is also effectively reduced, so that the grouping scheme and the precoding scheme can effectively explore the key technology of the stratospheric large-scale MIMO communication system.

The invention provides a joint measurement method based on a laser radar and a binocular visible light camera, which can obtain more accurate and dense three-dimensional information of a target simply and efficiently. In this joint measurement method, the lidar is assumed to be a camera device with fixed internal parameters, the 3D point cloud data is directly projected into a 2D image, and the image processing method is used to calibrate the lidar device and the dual The rotation and translation relationship between the cameras; among them, the idea of ​​finding the matrix norm and matrix trace is innovatively introduced to solve the rotation matrix; finally, the fusion of lidar and binocular stereo vision point cloud data can not only obtain accurate positions And attitude information, and can reconstruct the special texture and characteristic information of the target surface, which is very useful for the rendezvous and docking of spacecraft in the military field, the capture of hostile satellites, etc., as well as the workpiece measurement and unmanned driving in the civilian field. Value.

The invention belongs to the technical field of radar tracking system; similar systems, and discloses a method based on fuzzy C-mean point trace clustering, grouping measurement values; removing clutter; estimating the number of targets in the group; selecting the initial center; calculating the membership matrix U t ; Perform defuzzification; estimate the integrity of the cluster; update the clustering matrix U t+1 and the center of the cluster v i ; compare U with the matrix norm t and U t+1 ;if ||U t+1 -U t ||≤ε, stop. Otherwise, set t=t+1 to perform a new round of updating; finally defuzzify the measured values ​​according to the clustering matrix. In order to find the initial target center, the present invention takes into account the predicted localization and survey rate. At the same time, the integrity of the clustering is considered in the iterative process of the FCM algorithm. Compared with the traditional method, the present invention has better robustness and effectiveness, and can be used to correctly cluster the multi-moving targets detected by the radar, so as to track the targets better.

The invention relates to a system for calculating power supply abundance of a urban power network, which is characterized by comprising an equivalent model fitting module, a trust domain subproblem solution module, a target function real descend value solution module, a target function searching step-length searching direction adjusting module and an alternating current power flow calculating module; the network basic data and the calculated initial value of the city electric network are input into a calculating system; in the equivalent model fitting module, the equivalent second-order modelof the target function is output according to the basic data and calculated initial value; when the HESSEN matrix norm value of the equivalent second-order model is smaller than permissible error, namely the abundance value of power supply capability of the power network is output finally by the target function real descend value solution module. The system for calculating power supply abundance of the urban power network fully exerts the advantages of effectiveness and local superlinear convergence when the trust domain method is used for solving morbidity optimization problem, simultaneously the alternating current power flow algorithm reflects the influence of voltage level of the system, transmission capability of branches and reactive power flow to the power supply capability of the network precisely, and the voltage constraint of all nodes in the network can be calculated.

The invention discloses an energy efficiency optimization method suitable for a large-scale multiple-input-multiple-output system. The method comprises the steps: firstly creating an energy efficiency optimization model of a joint optimization base station antenna number, an antenna subset and transmission power; secondly proposing a low-complexity iterative search method, supposing that the base station antenna number is M, and carrying out the traversal of the base station antenna number from one to M; thirdly selecting one antenna number and then employing a suboptimal antenna selection algorithm based on a channel matrix norm for the selection of an antenna subset, wherein the energy efficiency is a quasi-concave function of transmission power after the antenna number and the antenna subset are determined, so the optimal transmission power can be solved through employing a convex optimization theory, and at this moment the energy efficiency corresponding to the optimal transmission power is the optimal energy efficiency under the current antenna number; finally comparing the energy efficiencies of M times, and obtaining the optimal energy efficiency of the system, the optimal antenna number, the antenna subset, and the transmission power. The method can reduce the cost of the system while improving the energy efficiency.

The invention discloses a beam structure damage identification method based on modal flexibility curvature matrix norms. The beam structure damage identification method includes the steps of conducting modal testing to obtain modal flexibility matrixes before and after beam structure damage respectively; acquiring curvatures of the flexibility matrixes before and after beam structure damage, sequentially acquiring the norms of column vectors of flexibility curvature matrixes and utilizing norm differences to position structure damage; computing a beam structure joint damage degree according to relative change of the flexibility curvature matrix norms so as to obtain a unit damage degree through computation. The beam structure damage identification method is capable of positioning single-damage and multi-damage working conditions of a beam structure effectively and capable of identifying the damage degree precisely, well overcomes the defect that existing modal flexibility curvature indexes cannot be used for damage degree identification, and can be applied to nondestructive testing and damage degree evaluation of the beam structure.

The invention discloses a representation method of human motion, which maps the human motion to a low-dimensional embedding space through nonlinear dimension reduction; models the data after dimension reduction with a linear time series model. Among them, the human motion is mapped to a low-dimensional embedded attitude space through nonlinear dimensionality reduction; the realization is: z=(x1+jy1, x2+jy2,...,xM+jyM)T. Model the dimensionally reduced data with a linear time series model; implementation steps: motion modeling based on a linear time series model; establish a p-order autoregressive model (AR) including; a p-order autoregressive model (AR) AR (p) has the following parameters: coefficient matrix Ak∈Rm×m, the parameter v introduced to ensure that the mean value of the dynamic process is non-zero, and the covariance matrix Q of Gaussian white noise. Let Ak be a diagonal matrix, then z(t) Each component of is independent; Given two autoregressive models (AR) A=[v, A1, A2,..., Ap] and A'=[v', A1', A2',..., Ap ’], then the distance metric D(A, A’)=‖A-A′∥F, where ‖·‖F represents the F-norm of the matrix.

The invention discloses a method for analyzing skiing motion sequences based on Hidden Markov, which belongs to the field of motion posture data processing and analysis. The realization method of the present invention is: by windowing the original attitude data of the skier collected by the sensor, dividing the original attitude data into data frames on a time series, and then using SVD and matrix norm to extract the "symbolic value" representing the attitude matrix . Based on Hidden Markov, the probability transition model of the skiing motion sequence is built, and the parameters of the probability transition model of the skiing motion sequence are optimized by using the "symbolic value" sequence on the time series, and the optimal path of the hidden state of the skiing motion sequence is calculated. Calculate the probability of each skiing state at the next moment by calculating the optimal hidden state path and the probability transition model, and then predict the skiing state at the next moment. Improve the training effect for ski trainers.

The method includes following steps: after decoding, acquires norm NDCT of DCT coefficient matrix of current macro block / sub block; calculates the transformed norm Nq of quantify matrix set by the current macro block / sub block; calculates relevant quantify step size of DCT coefficient matrix being relative to transformation of quantify matrix DCT, a=Nq / NDCT; compares a with experience-threshold k; if a<k, then ends the process. If a>k, then makes smoothing process for current macro block / sub block and their neighboring pixels. The unit used in video encode and decode system consists of DCT coefficient matrix norm calculating device, quantify matrix DCT transformation calculation device, relevant quantify step size calculation device, comparator and smooth-processing device.