39 results about "Nonlinear conjugate gradient method" patented technology

Disclosed is a blind restoration method for the motion blurred image; the steps are as follows: (1) the image is converted through the cepstrum method to figure out the blur extent and the blur direction of the blurred image; (2) the blur extent and the blur direction of the blurred image figured out in step (1), and the total variation (TV) method are adopted to process the restoration towards the blurred image. The method based on the total variation (TV) includes: the fixed-point iteration used as the outer-loop and the conjugate gradient method used as the inner-loop are adopted for loop iteration to obtain the restored image. The blind restoration method has the advantages of strong self-adaptation, strong anti-noise ability and good robustness; the blind restoration method has validity and practicality according to the processing effects of the simulation picture and the photographed picture.

The invention discloses a non-linear conjugate gradient three-dimensional inversion method of a magnetotelluric field, which comprises the following steps of: a, establishing a three-dimensional model, and determining model parameters; b, acquiring impedance response through the model resistivity by using the staggered-grid finite difference of the magnetotelluric field as a forward modeling method; c, comparing the impedance which is acquired by forward modeling with an actually measured impedance to acquire an impedance data deviation; d, judging whether a target function value is enough small or not, finishing iteration when the target function value is small enough, otherwise entering the next step; e, calculating the gradient of a target function, and acquiring a search step-length and a search direction by using a non-linear conjugate gradient method as an inversion method so as to calculate a model modifier, wherein the resistivity calculated by using the impedance data error and current iteration is used as a preprocessing factor in the inversion method; and f, modifying the model, and returning to the step b. The non-linear conjugate gradient three-dimensional inversion method of the magnetotelluric field has the characteristics of high efficiency and high accuracy.

The invention relates to a tuning calculation method of a distributed conjugated gradient method based on MPI. The calculation process of the conventional serial calculation method can not effectively utilize the advantages of a multi-kernel platform. When a new node is added to a calculation cluster, a pre-tuning algorithm is adopted to determine the calculation thread number of the node and obtain the optimal matrix block size suitable for cluster calculation; a matrix data structure is converted into a block compression structure according to the optimal block size; before calculation of the task to carry out, the calculation amount is allocated to the thread of each node according to the pre-tuning data of each node; and the calculation task is automatically allocated to the calculation node of the cluster by use of MPI when the calculation process of the conjugate gradient method executes the matrix-vector multiplication, and the calculation result is actively reduced to the main node. The method can lower the constant for calculating the time complexity and save the storage space by adopting a highly-compressed, flexible and convenient BCSR structure to segment a sparse matrix.

The invention discloses a rapid planar sparse array synthesis method. The method has the advantages that a conventional planar array synthesis constraint optimization model is converted into an unconstraint optimization problem by constructing a Lagrange function, so that calculation ill-posed problems in the iterative optimization process of a planar array are avoided, an array weight vector can be updated by the aid of closed-form solution in each time of iteration, planar array synthesis problems can be solved without an optimization tool, the method is more universal and transportability, an inverse matrix in the closed-form solution is quite large in scale as a two-dimensional space angle sample number of the planar array is increased in a quadratic manner, the inversion problem of the large-scale matrix is solved by introducing a conjugate gradient method, and convergence of planar array synthesis is accelerated, so that the method is higher in real-time performance and particularly applicable to occasions with high requirements on array optimization real-time performance and universality.

The invention provides an industrial process dynamic optimization system based on a nonlinear conjugate gradient method, which comprises an in-site intelligent meter, a DCS system and a host machine, wherein the in-site intelligent meter is connected with an industrial process object, the host machine comprises a restriction processing module, an initialization processing module, an ODE solving module, an iteration optimization module and a convergence judgment module, the restriction processing module is used for processing the control variable boundary restriction in the optimization process, the initialization processing module is used for setting initialization parameters, the OED solving module is used for solving ordinary differential equation groups of a dynamic optimization question, the iteration optimization module is used for searching a decision vector w which makes a target function J optimum, the convergence judgment module is used for judging whether the error absolute value of the target value obtained by the current convergence and the target value obtained by the former convergence is smaller than or equal to the set convergence precision omicron, and the current optimum vector w*, the current optimum target value J* and the current convergence time number k are stored if the error absolute value of the target value obtained by the current convergence and the target value obtained by the former convergence is smaller than or equal to the set convergence precision omicron. The invention also provides an industrial process dynamic optimization method based on the nonlinear conjugate gradient method. The invention can simultaneously meet the requirements of high efficiency and high precision of the on-line dynamic optimization solving.

The invention provides a hybrid precoding design method for a wideband large-scale MIMO system. The method comprises the following steps: a simulation precoding matrix is initialized, and a vector form of the simulation precoding matrix is obtained by means of a matrix straightening operation; simulation precoding is updated by means of a Riemann manifold-based conjugate gradient method; digital precoding is updated according to a least-mean-square-error criterion; and the above steps are iterated to update simulation precoding and digital precoding until the iterative computing process is convergent. According to the method, frequency-flat simulation precoding is designed for a frequency selective channel, signal to interference plus noise ratios of various users in an actual system can be effectively increased, the transmission rate performance of the actual system can be improved, and compared with traditional pure digital precoding, the performance approximate to pure digital precoding is obtained with less hardware loss.

The invention discloses a rotation intensity-modulated optimization method based on a dynamic grating and a radiotherapy device. Compared with the prior art, the method and the device have the advantages that the advantages of rapid optimization of IMRT Sliding Window and a tumor morphological feature algorithm are combined, and the shortcomings thereof are effectively overcome; for rotating field, results based on field intensity distribution and subfield splitting are used; after dynamic conversion, a conjugate gradient method is used to optimize the subfield weight, and a perturbation technology optimizes the subfield shape; the requirement of computing time is met; and requirement of effect optimization is met.

The invention provides a basic module-based mask main body graph optimization method. The method comprises the following steps of: constructing a mask main body graph into a plurality of superposed basic modules the single side size of which is greater than the threshold, namely the mask main body graph can be shown convolution of the basic modules and the coefficient matrix showing the positions of the basic modules; constructing the optimization target function F into squaring the Euler distance between the target graph and an image in photoresist corresponding to the current mask main body graph; and then based on an ABBE vector imaging model, optimizing the mask main body graph by adopting an improved conjugate gradient method. According to the method, the single side size of any part of the optimized mask main body graph is greater than a preset threshold can be automatically ensured in the mask optimization process. In addition, according to the method, only the mask main body graph is optimized without introduction of any auxiliary graphs, so that no auxiliary graphs excessively close to the main body graph can be produced. Therefore, the method can effectively improve the manufacturability of the optimized mask on the premise of improving the imaging quality of a photoetching system.

The invention provides a three-dimensional magnetotelluric tipper inversion method using nonlinear conjugate gradients and belongs to the technical field of magnetotelluric sounding. The method comprises steps as follows: S1, inputting data and performing ranking; S2, distributing the parameter of each frequency point to all cpus for tipper inversion and computing; S3, stopping parallel computing and combining the data, and comparing the difference between an invention result and the input date; S4, distributing the data difference and the parameter of each frequency point to all cpus for parallel computing of objective function gradients, computing the search direction p and the step length a and accordingly obtaining a model modification amount; S5, updating a module according to the model modification amount and stopping iteration when an objective function is one hour enough, otherwise, using a new model to repeat the steps from S2 to S5. According to the inversion method, tipper three-dimensional inversion is realized on the basis of tipper data by introducing a nonlinear conjugate gradient method, and the invention result approaches to real underground three-dimensional electrical structure information greatly; the method adopts a parallel computing structure, and the computing efficiency is improved.

The invention discloses a compressed sensing signal reconstruction method based on a conjugate gradient method. According to the invention, under the condition of known sensing matrixes and observed signals, the reconstruction of a source signal with sparsity under the white gaussian noise environment can be achieved by the conjugate gradient method. According to the compressed sensing signal reconstruction method, the deficiency that in the prior art the reconstruction method based on orthogonal matching pursuit is high in complexity and the deficiency that the prior art based on the gradient descent method is slow in rate of convergence and low in reconstruction accuracy are overcome, thus the source signal can be reconstructed with lower complexity and higher reconstruction accuracy in actual application.

The invention provides a supervised linear dimensionality reduction method with a separation probability of a minimax probability machine and belongs to the technical field of computer machine learning and statistical learning. The method comprises the steps that: a supervised linear dimensionality reduction model with the separation probability of the minimax probability machine is established, an input of the model is a sample set with multiple dimensions and categories, and an output is a projection matrix, when the dimensions are reduced to 1 dimension, an object belongs to a single projection vector object, when the dimensions are reduced to multiple dimensions, objects belong to multiple projection vector objects. According to the method, a separation probability between samples is used as a distance measurement between the categories, a conjugate gradient method is used for optimization, and finally, each category pair has a projection matrix of a maximum separation probabilityas far as possible. According to the method, the distinguishability of the data and the accuracy and efficiency of the subsequent classification can be improved, and a good application effect can be achieved in the problems of multiple types of dimension reduction.

The invention belongs to the radar foresight super-resolution imaging technical field, and discloses a radar foresight super-resolution imaging method based on the maximum entropy criterion; the method comprises the following steps: using dual-channel foresight scanning radar echo signals to modeling; introducing the entropy as target distribution prior information, and building a maximum posterior imaging model according to the Bayes criterion; employing an improved conjugate gradient method to solve the imaging model. The method combines with sum and difference channel processing so as to effectively alleviate single channel deconvolution problem inherent pathosis, and uses the entropy as the target prior information so as to prevent the echo prior information inaccurate problems, thus breaking the limits on the orientation resolution by the antenna aperture; the method uses the improved conjugate gradient method to carry out iteration solving for the foresight imaging result, is high in efficiency, thus realizing engineering high resolution foresight imaging.

The invention discloses a triple acceleration topology optimization method. According to the invention, by improving three aspects of multi-grid density mapping, the conjugate gradient method based onpre-processing and an initial value, and local updating, a design domain is divided into a plurality of layers of grids with different thicknesses through multi-grid density mapping, topological solving is conducted on a coarse grid layer firstly, an obtained optimization result is mapped to a next layer of fine grid layer to serve as an initial value, a topological optimization process on the fine grid layer skips redundant iteration, and the topological optimization process is accelerated; According to the conjugate gradient method based on preprocessing and an initial value, a preprocessing regulator and the initial value are added to iterative solution of a topological optimization equation, and the solution speed is increased; in the local updating, an update unit is selected, and the number of unit updates during iteration is reduced to achieve an acceleration effect. Based on the method, triple acceleration of topological optimization is realized, the topological optimization process is remarkably accelerated, and an optimization result with high precision and low calculation cost is obtained.

The invention discloses an unconstrained static structural analysis method based on Householder transformation. The unconstrained static structural analysis method includes the steps that step1, a structure rigid body displacement mode matrix X is built; step 2, six corresponding n*n order Householder matrixes Pi are built according to the rigid body displacement mode matrix X; step 3, the Householder matrixes Pi built in the step2 are used for performing orthogonal similar transformation on a structure original rigidity matrix to obtain a structure rigidity matrix Kp with a rigid body mode removed; step 4, a correction conjugate gradient method is adopted for solving the rigidity equation (KPUP=FP) which is obtained after removing the structure overall rigid body displacement mode by the Householder matrixes. The structure rigid body displacement mode is removed, and the appointed minimum error threshold value of the conjugate gradient method is controlled, so that structural response is accurately solved; calculation and implement steps are concise, and it is unnecessary to modify a finite element calculation frame commonly used currently; by the adoption of the correction conjugate gradient method, the sparse characteristic of the structure rigidity matrix can be used well, and due to step solution, the overall solving process is small in occupied space and high in calculation efficiency.

The invention discloses a large-scale MIMO precoding method based on a conjugate gradient method. The method comprises the following steps of (1) estimating a channel matrix, and computing an RZF precoding expression through the estimated channel matrix; and (2) substituting the RZF precoding matrix into a transmission signal expression, and arranging to a form of a linear equation set AT=s; and solving the linear equation set by adopting the conjugate gradient method, and then obtaining the transmission signal expression. According to the method, an inverse matrix solving process is convertedinto multiplication of the matrix and the vector, so that the computation complexity is reduced, the improved conjugate gradient method optimizes an initial value of the conjugate gradient method according to the channel characteristics of a large-scale MIMO matrix, the convergence speed of the conjugate gradient method is accelerated, after two iterations, when the large-scale MIMO precoding based on the conjugate gradient method is 16 dB, a bit error rate reaches 10<-6>, and furthermore, after the improved conjugate gradient method carries out two iterations, the bit error rate of the method is close to the bit error rate of an RZF precoding algorithm.

The invention discloses an inaccurate Newton's solution-based block adjustment method employing a preconditioned conjugate gradient method. The preconditioned conjugate gradient method is introduced into a traditional block adjustment flow to solve a large-scale normal equation; and storage and direct inverse operation for the normal equation are avoided. According to the method, a normal equation coefficient matrix B is not directly stored while a matrix-vector product normal equation coefficient matrix Bd is calculated in a point-by-point manner; the block adjustment flow is suitable for a parallel design scheme; meanwhile, an accurate solution of the normal equation is replaced with an inaccurate Newton's solution, so that the number of iterations of the preconditioned conjugate gradient method is reduced; the memory capacity of data is reduced; and the determining speed of the overall block adjustment is improved.

The invention discloses a low-noise least square reverse time migration method. The method comprises the following steps: establishing a least square reverse time migration function; a sparse constraint term is added to a least square reverse time migration function to suppress noise, and a target function is obtained after sparse constraint of the Cauchy constraint term; and determining parameters of an objective function of the least square reverse time migration function after sparse constraint of the Cauchy constraint term, optimizing the gradient of the objective function, and performing minimization solution on the least square reverse time migration function by adopting a precondition nonlinear conjugate gradient method. According to the low-noise least square reverse time migration method disclosed by the invention, a new method for determining regularization parameters and hyper-parameters in Cauchy terms is provided, a better imaging result can be obtained, the convergence speed of inversion can be increased, and a method for determining step length is provided for an objective function containing Cauchy constraint terms.

The invention provides a gravity gradient data joint inversion method, relates to the technical field of geophysical inversion, and aims to introduce a gradient depth weighting function, physical property constraints based on pair/index transformation and other methods on the basis of focusing inversion in order to solve the problems of insufficient spatial resolution and the like of gravity gradient data inversion at present. Six components are combined in the full-tensor gradient data, and three-dimensional density inversion is achieved by using a nonlinear conjugate gradient method. Meanwhile, in order to improve the visualization effect and operability of inversion, a software platform with a visualization function is developed on the basis of Python language, PyQt, Matplotlib and other toolkits as proposed inversion methods, and the distinguishing capability of near geologic bodies and the longitudinal space imaging capability of target bodies are effectively improved. A softwareplatform and a development method thereof have the advantages of being easy to use, practical and the like.