32 results about "Sparse learning" patented technology

The invention relates to a broadband planar array antenna light-operated beam forming method based on multi-task sparse learning, and the method comprises the steps: deducing an array factor expression of a broadband array as priori knowledge through employing a rectangular planar array antenna; constructing a multi-frequency-point beam joint optimization model of the planar array antenna based on the neural network; based on the priori knowledge, calculating a total loss function of the model by using expected broadband array pattern data and a group inhaul cable regularization method, and training a multi-frequency-spot beam joint optimization model; and based on the trained model, obtaining the optimal array layout and the amplitude and delay distribution of each array element, and calculating to obtain a low-sidelobe directional diagram in which the deflection angle is irrelevant to the frequency. According to the invention, the deflection angle of the array radiation beam is kept constant in a broadband range, and the problem of'beam squint 'of a traditional phased array is solved; and meanwhile, the beam sidelobe level of each frequency point is effectively suppressed, and the grating lobe problem of the uniform array on the broadband is avoided.

The invention discloses a power distribution room remote control inspection and intelligent early warning method based on the Internet of Things, and the method comprises the following steps: arranging multiple types of sensors in a power distribution room to collect environment and equipment data, carrying out the synchronization, filtering and normalization of an edge calculation unit, extracting features through a lightweight artificial intelligence model, and detecting the abnormality; the method comprises the following steps of: uploading to a cloud end through a 4G and LoRa dual-mode network, performing time sequence prediction in a long-short-term memory network containing physical constraint gating, and calculating the fault risk of each monitoring point in combination with a Beta-Bernoulli priori sparse learning dynamic optimization model structure to generate a multi-stage early warning and control instruction so as to realize cloud-side collaborative intelligent inspection and remote control. According to the invention, through fusion of edge lightweight intelligent analysis, cloud physical constraint gating and Bayesian sparse learning, real-time sensing, time sequence prediction and multi-level intelligent early warning control of the operation state of the power distribution room are realized.

The invention relates to the related technical field of photovoltaic power prediction optimization, in particular to a photovoltaic power prediction optimization method. According to the photovoltaic power prediction optimization method, firstly, modal decomposition is performed on time sequence preprocessing, then, triangular topology aggregation optimization is utilized, and then, a prediction model and an optimization method are determined; the method mainly determines a time domain convolutional network for time sequence prediction, frequency-time sparse learning in time sequence prediction and a Transformer encoder for time sequence prediction, and learnable parameters of TCN, FreTS and Transformer encoders, the method enhances the ability of a prediction model to extract meaningful features and improves the prediction precision, and by integrating the TCN, FreTS and Transformer encoders, the prediction accuracy is improved, and the prediction efficiency is improved. The model constructs a unified prediction framework, and the framework considers local feature extraction, frequency-based optimization and global dependency relationship learning. By means of the synergistic effect, the model can achieve high precision, high noise immunity and adaptability to various prediction tasks.

In various examples, a radar system includes a logic circuit with a memory array for processing radar reflection signals. In a specific example, a method includes generating output data indicative of the reflection signals' amplitudes, and discerning angle-of-arrival information for the output data for the output data by correlating the output data with an iteratively-refined estimate of a sparse spectrum support vector (“support vector”). The estimate includes: iteratively updating a set of parameters associated with previous values of the support vector including a covariance estimate, and a statistical expectation among a plurality of support vectors; and pruning, for each iterative update, certain of the plurality of support vectors having amplitudes which are insignificant relative to the statistical expectation of the support vector of in a preceding iteration.

The invention discloses a radial power distribution network topology identification method, system and device and a medium. The method comprises the following steps: acquiring node voltage amplitude measurement data; calculating a sample covariance matrix based on the voltage amplitude measurement data to obtain the covariance of the voltage amplitude between the nodes; estimating a precision matrix of the covariance matrix through a graph lasso model; calculating a partial correlation matrix based on the precision matrix, wherein the partial correlation matrix is used for representing a direct correlation coefficient between nodes in the radial power distribution network; and deriving an adjacent matrix through a maximum spanning tree based on the partial correlation matrix to obtain a radial power distribution network topological structure. According to the method, sparse learning and partial correlation analysis are combined to deal with the problem of limited data volume; the key condition dependency relationship between the nodes can be quickly and accurately identified only through voltage amplitude measurement data without a large number of samples and phase angle information, and the identification precision is high.

The invention discloses an ultrasonic cross-metal channel estimation method based on improved block sparse Bayesian learning, and belongs to the field of ultrasonic cross-metal communication channel estimation. Constructing a pilot signal relationship between the receiving end and the transmitting end on the two sides of the metal wall; dividing a to-be-estimated metal channel impulse response into zero-value blocks or non-zero blocks of block sparse Bayesian according to whether the to-be-estimated metal channel impulse response is zero or not; the method comprises the following steps: iteratively updating hyper-parameters a, b, Bi and lambda of a non-zero block by adopting an expectation maximization algorithm, realizing structured sparse learning, calculating a maximum posteriori estimation result by utilizing a final hyper-parameter of the non-zero block, and taking the maximum posteriori estimation result as an estimation value of ultrasonic cross-metal channel pulse response. According to the method, a block sparse Bayesian learning algorithm is improved by utilizing the characteristics of metal channel pulse response, and compared with an original algorithm, the accuracy of channel estimation in a low signal-to-noise ratio environment is improved.

A fault diagnosis method for a transmission chain based on joint entropy enhanced sparse learning using a zero sequence current includes the following steps: data acquisition and preprocessing; establishment of a rotating machinery fault diagnosis model for sparse feature learning of a zero sequence current; and obtaining of a diagnosis result by inputting the preprocessed zero sequence current data to the trained rotating machinery fault diagnosis model. The fault diagnosis method for a transmission chain based on joint entropy enhanced sparse learning using a zero sequence current can extract a weak fault feature in a current signal automatically and efficiently without relying on traditional signal processing techniques and diagnosis experience, and has good robustness for signals containing noise.

The present invention relates to the technical field of machine learning, and particularly to a miRNA-disease prediction method based on sparse learning and random walk, including performing sparse learning on a disease-miRNA association matrix; reconstructing a new disease-miRNA association matrix after removing the sparse part, and obtaining an initial probability matrix through the reconstructed association matrix; constructing a disease integrated similarity network and a miRNA integrated similarity network according to disease similarity and disease Gaussian kernel similarity; integrating the reconstructed association networks into a heterogeneous network; performing multi-level random walk for different purposes by using the association information between network nodes; and effectively predicting the potential similarity between diseases and miRNAs. The present invention fully considers the influence of the initial probability node information in the restart random walk algorithm on the model performance, and performs potential association prediction on the basis of capturing the topological structure characteristics of the similarity network by using the multi-level random walk algorithm.

The application discloses a scanning radar two-dimensional super-resolution imaging method based on Bayesian sparse learning, first constructs a bearing-pitch two-dimensional scanning radar signal model, then, according to a maximum posteriori criterion under a Bayesian framework, establishes a sparse optimization target function about target scattering and environmental noise, finally, utilizes a conjugate gradient algorithm and Kronecker product properties to accelerate iterative estimation of target scattering and noise power, realizes adaptive sparse two-dimensional super-resolution imaging of the scanning radar. The method solves the problems of high complexity and poor noise robustness of prior art means, compared with prior art two-dimensional super-resolution methods, has lower calculation complexity, is more robust, has excellent noise adaptive capacity, and can realize two-dimensional scanning radar super-resolution imaging under a low signal-to-noise ratio condition.

The invention discloses a transmission chain cross-domain migration fault diagnosis method based on sparse learning and Riemannian measurement. The method comprises the following steps: data acquisition and preprocessing; constructing a feature extraction model based on a sparse auto-encoder; the extracted features are processed through covariance mapping and Riemannian dimension reduction; and inputting the manifold features after dimension reduction into a multi-layer perceptron classifier, and classifying fault types. The invention provides a feature extraction model based on a sparse auto-encoder, the model can automatically extract invariant features under various working conditions, and in combination with class discriminant analysis and a source domain-target domain matching strategy, effective fault identification and diagnosis precision improvement among different working conditions are realized.

The application provides an improved sparse learning out-of-grid direction-of-arrival estimation method based on variational Bayesian. The method is characterized in that: the original data received by a hydrophone array is preprocessed, a real value transformation is used to convert a vectorized covariance matrix signal in a complex number field to a real number field, and the idea of variational sparse Bayesian learning and grid evolution is combined to make the grid evolve from an initial uniform grid to a non-uniform grid adaptively in an iteration process. The evolution process includes grid updating and grid fission. The evolved grid points are gradually close to the real source position through the alternately iterative grid updating process and grid fission process. Compared with the traditional compressed sensing method, the method has higher DOA estimation accuracy, reduces the operation complexity, optimizes the operation efficiency, improves the resolution capacity of the source, and has higher application value in actual engineering, especially in the case of few snapshots and low signal-to-noise ratio.

The invention relates to an individual specific channel selection method for an SSVEP brain-computer interface system. The method comprises the following steps: firstly, quantitatively table-verifying the importance of each channel based on task-related spatial filtering; secondly, through channel sparse learning based on spatial distance constraint, the importance difference between the SSVEP task related channel and the unrelated channel is maximized, and the importance of the unrelated channel tends to 0; and finally, based on the significant difference of the importance degrees between the channels, adaptively selecting a proper individual specific channel. According to the method, the SSVEP sensitive channel can be accurately captured, individual specificity knowledge is fully utilized, and the individual specificity SSVEP-BCI performance is further improved.

The invention relates to a graph neural network node classification method based on mutual exclusion group sparse learning, and the method comprises the following steps: introducing a mutual exclusion group sparse regularization mechanism into an existing graph neural network, and finely regulating and controlling a feature selection process through a strategy of intra-group feature sharing and inter-group feature competition; the regular term is embedded into the graph neural network, and according to the feature importance and the noise level of different network layers, the weight of feature retention and elimination and the proportion of balance group sparse and mutual exclusion sparse regularization are adaptively adjusted, so that the robustness and the stability of feature selection are enhanced; and fine control and selection of multi-scale image features are realized. In practice, by using the method, the flexibility of feature selection and the network generalization of the neural network can be improved, and meanwhile, the robustness and the expression ability of the model are remarkably enhanced.

This invention discloses a fault diagnosis method for cross-domain migration of transmission chains based on sparse learning and Riemannian metric, comprising the following steps: data acquisition and preprocessing; constructing a feature extraction model based on a sparse autoencoder; processing the extracted features through covariance mapping and Riemannian dimensionality reduction; and inputting the dimensionality-reduced manifold features into a multilayer perceptron classifier to classify the fault type. This invention proposes a feature extraction model based on a sparse autoencoder, which can automatically extract invariant features under various operating conditions and, combined with class discriminant analysis and source-target domain matching strategies, effectively improves the accuracy of fault identification and diagnosis across different operating conditions.

The invention discloses an equipment dynamic feature selection method and system based on matrix Sketch, and relates to the technical field of equipment feature data processing. The method comprises the following steps of: when equipment state characteristic data is updated, performing data insertion on an initialization matrix Sketch; executing a space release operation and updating the matrix Sketch; performing sparse learning feature selection according to a sparse feature selection algorithm; and the sparse learning feature selection frequency is adjusted according to the dynamic feature selection frequency analysis and evaluation result. According to the method, sparse learning is carried out on the matrix Sketch, compared with sparse learning on original data, the space of the matrix Sketch is small, so that the calculation overhead of sparse learning can be greatly reduced, and the effects of improving the calculation efficiency of dynamic feature selection and improving the applicability are achieved; the problem of insufficient applicability caused by low calculation efficiency of dynamic feature selection according to network dynamic environment change in the prior art is solved.

The invention discloses a feature selection and dimension reduction method and system for multi-omics data fusion, and the system comprises a multi-omics data collection and heterogeneous preprocessing module which is used for carrying out the omics specific preprocessing of genome, transcriptome, proteome and metabolome data; the cross-omics heterogeneous graph neural network fusion module is used for constructing a heterogeneous graph and carrying out information fusion by adopting a heterogeneous graph attention network; the feature selection module based on multi-task sparse learning is used for selecting a key feature subset from the fused features; a non-linear dimension reduction module based on an adversarial variational auto-encoder and used for performing dimension reduction on the key feature subset to a low-dimensional submerged space; and a result visualization and biological interpretation module. According to the method, efficient fusion and dimension reduction of heterogeneous and high-dimensional multi-omics data are realized, the accuracy of feature selection and the interpretability of results are improved, and an effective calculation tool is provided for disease typing and biomarker discovery.

The invention relates to the technical field of signal denoising, and discloses an EWT-TQWT-AFSL-based joint denoising method and system, and the method comprises the steps: collecting an original ultrasonic echo signal of a transformer winding, dividing the original ultrasonic echo signal into a plurality of adaptive frequency bands through EWT, and generating an intrinsic mode component group comprising a high-frequency component and a low-frequency residual component; performing adjustable Q-value wavelet transform TQWT on the high-frequency component to obtain a multi-layer wavelet coefficient set, and performing soft threshold de-noising on each layer of coefficient by adopting an adaptive threshold function based on noise variance estimation; performing fusion reconstruction on the denoised high-frequency component, the unprocessed low-frequency residual component and other reserved components to generate an intermediate signal; and constructing an adaptive feedback sparse learning AFSL model, inputting the intermediate signal, solving a sparse reconstruction optimization problem through iterative weighting and feedback updating, and outputting a final de-noised signal, thereby effectively solving the technical problem that a traditional de-noising method is difficult to give consideration to both noise suppression and signal fidelity in a complex noise environment.

The application discloses a device dynamic feature selection method and system based on a matrix Sketch, and relates to the technical field of device feature data processing. The method comprises the following steps: inserting data into an initialized matrix Sketch when device state feature data is updated; performing a space release operation and updating the matrix Sketch; performing sparse learning feature selection according to a sparse feature selection algorithm; and adjusting the sparse learning feature selection frequency according to a dynamic feature selection frequency analysis evaluation result. The application performs sparse learning on the matrix Sketch. Compared with sparse learning on original data, the calculation overhead of sparse learning can be greatly reduced due to the small space of the matrix Sketch, the calculation efficiency of dynamic feature selection is improved, and the applicability is improved, so that the problem of low calculation efficiency of dynamic feature selection according to network dynamic environment changes in the prior art and the resulting insufficient applicability are solved.

The application discloses a robust graph convolutional network method based on space-time sparse learning. The method realizes spatial sparsity on each node through a TopK function, and proposes an attention mechanism based on time sparsity, that is, different weights are assigned to each dimension of a feature space according to different activation frequencies. The application provides an improved graph convolutional neural network, which has high robustness while maintaining the original network accuracy, and improves the anti-interference ability of the model in the face of noise.

The invention relates to the technical field of image processing, in particular to a depth interpretable multi-modal remote sensing image change detection method based on low-rank sparse learning, which comprises the following steps: constructing a training sample set; constructing a remote sensing image low-rank sparse semantic-modal coupling normal form, designing a modal conversion objective function and a change detection objective function, and performing iterative solution by adopting an alternating direction multiplier algorithm to realize alternating optimization; constructing a modal conversion sub-model, performing explicit decoupling on modal information and spatial semantic information, and performing cross reconstruction of a multi-modal image to obtain a remote sensing image of the same modal; constructing a change detection sub-model, and decoupling the dual-time-phase image to obtain difference characteristics of remote sensing images of the same mode; training the remote sensing image low-rank sparse semantic-modal coupling model by using the training sample set to obtain a trained model; and inputting a to-be-detected multi-modal remote sensing image into the trained model for change detection to obtain a change result of the to-be-detected multi-modal remote sensing image.

The application discloses a multi-label feature selection method based on sparse learning coupled mutual information, and comprises the following steps: inputting a feature matrix X, a label matrix Y and hyperparameters alpha, beta, gamma and delta, selecting a feature number k, and initializing a label correlation matrix Z and a feature correlation matrix W; calculating a diagonal matrix A and a similarity matrix S according to the feature matrix X, and calculating a graph Laplacian similarity matrix L of the feature matrix X x ; updating the label correlation matrix Z and the feature correlation matrix W through a target function, iterating n times, and obtaining an updated label correlation matrix Z n and an updated feature correlation matrix W n after the target function reaches a convergence condition; and obtaining the selected k features according to the 2-norm of W n . The application adopts the multi-label feature selection method based on sparse learning coupled mutual information, and the effectiveness of the method is proved by experiments; and the method simultaneously solves the suboptimal solution problem caused by a random initialization strategy in the sparse learning method.

The present invention proposes a hierarchical feature selection method and system based on label correlation and instance correlation. The feature selection task is decomposed into several sub-tasks, and a feature subset is selected for each sub-task to improve classification accuracy. The specific steps are as follows: S1, input the hierarchical training data set; S2, use a hierarchical tree structure to describe the semantic relationship between categories in the hierarchical training data set, and each internal node corresponds to a sub-task; S3, construct the objective function of the sub-task corresponding to the internal node using a loss term and regularization terms including sparse learning, label correlation, and instance correlation, optimize the weight matrix of each internal node according to the objective function, sort the weight matrices of each sub-task in descending order, select the features with large weight values, and complete the hierarchical feature selection task. The present invention not only considers the hierarchical relationship between categories but also utilizes the similarity between instances, thereby being able to more effectively select features and improve classification performance.

The application discloses a power distribution room remote control inspection and intelligent early warning method based on an Internet of Things, which comprises the following steps: collecting environment and equipment data by arranging multiple types of sensors in a power distribution room, extracting features and detecting abnormalities by a lightweight artificial intelligence model after synchronization, filtering and normalization by an edge computing unit, uploading to the cloud through a 4G and LoRa dual-mode network, performing time series prediction in a long short-term memory network containing a physical constraint gate, dynamically optimizing the model structure in combination with a Beta-Bernoulli prior sparse learning, calculating the fault risk of each monitoring point to generate multi-level early warning and control instructions, and realizing intelligent inspection and remote control in cloud-edge collaboration. Through the fusion of edge lightweight intelligent analysis, cloud physical constraint gating and Bayesian sparse learning, the application realizes real-time sensing, time series prediction and multi-level intelligent early warning control of the operation state of the power distribution room.

The invention provides a bearing fault diagnosis method based on current signal sparse learning. Comprising the following steps: step 1, collecting current signal data during stable operation of a motor and preprocessing the current signal data; step 2, performing analytic envelope transformation on the preprocessed current signal to obtain sparse representation in a frequency domain; 3, covering a part of frequency domain range by using a truncated off-network model, and jointly recovering sparse vectors through variational Bayesian inference (VBI); 4, optimizing the positions of the grid points by using the arithmetic sparse structure of the bearing fault characteristic frequency, and improving the accuracy of frequency estimation; step 5, iteratively updating the parameters through VBI until convergence, and obtaining an estimated value of the fault characteristic frequency; and step 6, comparing the obtained fault characteristic frequency estimation value with a theoretical value, and judging whether the bearing in the motor has a fault or not and judging the fault type of the bearing. According to the invention, effective diagnosis of different bearing fault types is realized. And the fault detection cost is effectively reduced.

The application discloses a kind of oilfield production robust prediction method and system based on neural additive model, it is related to petroleum well production prediction technical field, including: based on neural additive model SMART, by obtaining the input data consisting of multidimensional time series data, model training is carried out, and prediction model is constructed;Based on prediction model, using sparse learning strategy, mode-based measurement method and non-convex optimization algorithm, model optimization is carried out, and the production of oilfield is predicted according to the optimized prediction model.The application combines neural network and additive model, introduces mode-based measurement, sparse learning and non-convex optimization algorithm, effectively improves the accuracy, robustness, interpretability and efficiency of oilfield production prediction method, so that it is more suitable for application in actual scene.

The present invention discloses a server heat dissipation control system, method, storage medium and program product, which relates to the technical field of hardware heat dissipation. It includes a data acquisition layer, a resource management layer, a data reasoning layer and a decision execution layer. It can calculate the current sparsity of a neural network model through the multi-modal data and system load data of the server, then calculate the weight matrix of the neural network model through the current sparsity, use the neural network model and the weight matrix to reason about the multi-modal data, and finally generate a control signal based on the reasoning result to control the server to perform a heat dissipation action. Through the technical fusion of dynamic sparse learning and multi-modal fusion, high-performance heat dissipation control is achieved, solving the technical problem in the related art that there is a conflict between model resource occupancy and real-time performance during server heat dissipation, resulting in the inability to balance the requirements of control accuracy and real-time performance, and achieving the technical effects of optimizing sparse calculation and solving the conflict between multi-modal and embedded resources.

The application discloses a new energy power distribution network topology dynamic identification and abnormality detection method based on physical guidance sparse learning, and belongs to the technical field of power system automation. The method collects high-frequency measurement data of nodes of a power distribution network, and retains power fluctuation and negative value characteristics; a physical guidance neural network model embedded with linearized power flow equations is constructed to establish a mapping relationship from power to voltage; a hybrid loss function containing physical consistency error and topology sparsity regularization term is defined, the model is automatically trained under the condition of no topology label, and a parameter matrix is converged to a real sparse connection mode; based on the converged parameter matrix, the physical topology of the power distribution network is reconstructed, and voltage reconstruction residual error is calculated by using the model to realize accurate positioning of node abnormalities. The application does not need manual labeling and accurate physical parameters, has strong physical interpretability, can actively utilize new energy fluctuation to improve identification accuracy, and realizes integrated panoramic perception of topology dynamic identification and abnormality detection.

The technology described herein provides an improved framework for a face editing task performed by a machine-learning model. The technology provides a self-training strategy aimed at achieving more robust and generalizable face video editing. The self-training strategy helps overcome a shortage of training data relevant to the face editing task. The technology also provides a semantically disentangled architecture capable of catering to a diverse range of editing requirements. The technology also provides sparse learning to avoid over editing. The sparse learning technology partitions the model being trained according to facial regions being edited. This strategy teaches the model to transform only the most pertinent facial areas for a specific task. For example, when changing the eyebrows on a face the eye area will change, but the mouth area should remain unchanged.

The technology described herein provides an improved framework for a face editing task performed by a machine-learning model. The technology provides a self-training strategy aimed at achieving more robust and generalizable face video editing. The self-training strategy helps overcome a shortage of training data relevant to the face editing task. The technology also provides a semantically disentangled architecture capable of catering to a diverse range of editing requirements. The technology also provides sparse learning to avoid over editing. The sparse learning technology partitions the model being trained according to facial regions being edited. This strategy teaches the model to transform only the most pertinent facial areas for a specific task. For example, when changing the eyebrows on a face the eye area will change, but the mouth area should remain unchanged.

The present invention relates to the field of visual target tracking, and specifically provides a robust visual target tracking algorithm from coarse to fine based on sparse learning, including step S1: establishing a target appearance model, and S2: tracking the target. In step S1, context constraint and lasso regression are introduced, and the sparsity of the target response is embedded into the tracking framework; in step S2, a coarse-to-fine positioning method is adopted, and interactive Kalman filtering is used to supervise and participate in the positioning of the target, and multi-peak detection is performed on the response map during fine positioning. The target tracking algorithm provided by the present invention can effectively suppress the influence of internal and external interferences of the target by establishing a robust and sparse target appearance model, effectively avoid the occurrence of model drift or tracking failure caused by accidental peaks, and can improve the accuracy of positioning.