14results about How to "Improve clustering effect" patented technology

The invention relates to the technical field of multi-view clustering analysis, and discloses an incomplete multi-view clustering method and system, a storage medium and equipment. In order to solve the problems that a small amount of easy-to-obtain supervision information is not utilized in an existing method, and the clustering performance is deteriorated due to low view interpolation quality under high missing degree, the invention provides a technical scheme of combining paired constraint weighted interpolation and double-level feature fusion. The method comprises the following steps: firstly, screening similar samples by using pairwise constraint information, and reconstructing a missing view through similarity weighting; then extracting feature mean values and standard deviations of all views based on an encoder, aligning features through view hierarchy self-adaptive comparison learning and reserving private information, and optimizing feature distribution in combination with sample hierarchy semi-supervised loss; and finally, completing clustering through a Gaussian mixture model. According to the method, the pairwise constraint information is effectively utilized, the view recovery quality and the feature complementarity under the high-missing scene are improved, the clustering performance is remarkably improved, and the method is suitable for scenes such as data analysis of multi-view data missing and the like.

This application provides an image processing method, apparatus, computer-readable medium, and computer device. The image processing method includes: acquiring multiple original images and descriptive text corresponding to each of the multiple original images; converting the descriptive text corresponding to each of the multiple original images into text vectors to obtain text vectors corresponding to each of the multiple original images; performing clustering processing based on the text vectors corresponding to each of the multiple original images to obtain multiple original image clusters, and generating text embedding vectors corresponding to each original image cluster; using the text embedding vectors corresponding to each original image cluster as text-guided features, performing semantic editing processing on the original images in each original image cluster to obtain processed images, wherein the processed images are used as sample images for a machine learning model. The technical solution of this application embodiment can enrich the diversity of training datasets, thereby improving the generalization ability and robustness of the model.

This disclosure provides a matching model training method, device, and medium, relating to the field of computer technology. The method includes: clustering the original data in an unlabeled target dataset using a clustering algorithm to obtain multiple data clusters. The target dataset is obtained by combining two unlabeled candidate datasets. Each pair of original data from the multiple data clusters is concatenated to obtain N concatenated data. Then, each pair of concatenated data from the N concatenated data is combined to obtain K data groups. The label value of each data group is determined based on the clustering of the original data within each data group. An untrained matching model is trained based on the K data groups and the label values ​​of each data group until a trained matching model is obtained. Labels are added to the unlabeled data through clustering and data sources, and the effect is optimized through an iterative training process, ultimately resulting in a more accurate trained matching model.

This invention provides a deep multi-view clustering method based on multi-granularity contrastive learning, belonging to the field of computer vision technology. It solves the technical problem that single-granularity feature representations are insufficient to comprehensively capture the multi-level semantic information of data and the quality differences between different views. The method includes the following steps: S10, designing an independent encoder for each view to extract latent embedding features; S20, mapping the attention-enhanced features of each view to the same semantic space; S30, dynamically assigning weights based on the similarity between the features of each view and the global representation; S40, granular-level contrastive learning can enhance the compactness of the cluster structure within the granular sphere and align the granular sphere structure between cross-view granular spheres; S50, performing K-means clustering on the final global features to obtain the final prediction result. This invention enhances the weight of key features by adding a feature attention enhancement module to the view features.

This invention discloses a multimodal brain imaging data classification method and system based on a dual-graph autoencoder. The method includes the following steps: Step S01. Acquire multimodal brain medical imaging data of different target individuals and extract standard uptake ratio data and similarity matrix data; Step S02. Construct individual brain region maps and population maps; Step S03. Extract individual embedding features using an individual graph autoencoder and extract population embedding features using a population graph autoencoder; Step S04. Use the standard uptake ratio, brain region similarity matrix, individual embedding features, and population embedding features as views for enhanced multi-view subspace clustering; Step S05. Select key brain regions and discriminative brain region connections to participate in the next round of clustering as enhanced views until a preset convergence condition is met. This invention can achieve high-precision, highly interpretable data classification and simultaneously output biologically interpretable discriminative brain regions and connections.

The application discloses a radar signal processing method based on TOA sequence correlation degree, which comprises the following steps: parameter initialization; obtaining a radar signal pulse description word; dividing the radar signal pulse description word by using a grid method; extracting grids with a pulse number higher than a threshold value, and traversing all the extracted grids to extract the features of all the pulses contained in the grids in the TOA dimension, and the features are described by using TOA sequence feature vectors to obtain the TOA sequence feature vectors corresponding to each grid; quantifying the correlation degree between the TOA sequence feature vectors extracted by different grids to obtain corresponding correlation coefficients; performing grid merging and clustering, and outputting a radar signal clustering result. The method has high accuracy when clustering a multi-parameter agile periodic scanning radar, and is not affected by the pulse interval modulation mode of the radar. In addition, when facing a multi-parameter agile radar, the method can effectively prevent the problem that a traditional clustering algorithm is prone to produce an increased batch result.

The application discloses a spatial transcriptome data clustering method, comprising the following steps: firstly, preprocessing the original data, and constructing a spot space neighbor network; secondly, further learning spatial information and low-dimensional potential representation of gene expression through a graph attention automatic encoder, learning self-expression coefficient matrixes of different layers in the encoder through a multi-scale self-expression module and fusing the same together, adopting spectral clustering in a deep subspace clustering module, and feeding back the clustering label to a self-supervision module; the self-supervision module guides the learning of the potential representation by constructing a self-supervision path returning to the encoder; finally, biological analysis is carried out; the application learns low-dimensional potential embedding by integrating spatial information and gene expression profiles, and improves the clustering performance of spatial resolution transcriptome by using STMSGAL.

A millimeter wave radar-IMU pose parameter calibration method, comprising: collecting multiple sets of IMU and radar point cloud data on a grid plate, and accumulating multiple frames of point clouds in front and back; for point clouds with unchanged IMU attitude or position, point cloud registration is performed relative to the origin, the relative pose of the radar is calculated, the corresponding IMU data is obtained, and the relative pose of the IMU is calculated; based on the degenerated hand-eye calibration algorithm, a least squares loss function is constructed for the two types of relative poses in turn, and the initial values of the rotation and translation parameters are optimized; the point clouds with changing IMU poses are projected to multiple centers, the center point clouds are characterized by four-dimensional normal distribution, a re-projection error loss function of points and distribution characteristics is constructed, and the final parameter values are optimized. It solves the problem of missing IMU odometry or difficulty in fully rotating the body during calibration, and also proposes a millimeter wave radar point cloud registration method, which improves the adaptability of the millimeter wave radar-IMU system in sensor parameter calibration.

This invention designs an image classification method based on machine learning and spectral clustering of grains, comprising: acquiring an image dataset; extracting the feature vector of each image in the image dataset using a feature extraction model; creating an initial grain as a whole by treating the feature vectors of all images in the image dataset, and calculating the variable sparsity measure (VSM) value of the initial grain; if the VSM value is lower than a preset threshold, dividing the grain into two smaller grains using a 2-means clustering algorithm; continuing until the VSM values ​​of all grains meet the condition; calculating the similarity between the generated grains based on the center feature vector and the radius of the grains; dividing the generated grains into multiple clusters using spectral clustering based on the similarity between the grains, where each cluster represents a category; calculating the similarity between the feature vector of the image sample to be tested and the center feature vector of each grain, and taking the category corresponding to the cluster of the most similar grain as the classification result of the image sample to be tested. This invention provides more reliable and efficient technical support for the field of image classification.

The application provides a kmeans log classification method and device based on fusion of cut-off distance, and relates to the field of big data.The method comprises the following steps: selecting log samples with cosine distance within a first distance range from a center from a log sample set to form a first sample set, and selecting a first log sample from the first sample set according to sample density and adding the first log sample to a center set; selecting log samples with cosine distance within a Kth distance range from the first log sample from the log sample set to form a Kth sample set according to the value of K, and selecting a Kth log sample from the Kth sample set according to sample density and adding the Kth log sample to the center set, wherein the value of K is [1, K] in turn; and fusing the log samples in the center set until the number of the log samples in the center set is K, so as to serve as initial centroids of kmeans clustering.The scheme improves the clustering effect of the Kmeans algorithm.

The application relates to the computer technical field and provides a data clustering method, a data clustering system, electronic equipment and a storage medium, the method comprises the following steps: initializing an original corpus based on a clustering center point and user features to obtain a first target corpus; similarity matrixes suitable for the term frequency inverse document matrix are calculated according to segmentation adopted in term frequency statistics in the process of constructing the term frequency inverse document matrix; the similarity matrixes are input into the first target corpus, the vector cosine similarity of each query term and all non-query terms is calculated and arranged in descending order to obtain a recommended result of an extended word; a sorting result is obtained by combining the recommended result and user information; the first target corpus is updated based on the sorting result to obtain a second target corpus. The application realizes quick clustering of new data of the same type, improves the clustering efficiency and precision, solves the problem of too many iteration numbers caused by randomly selecting a clustering center, and improves the clustering effect of data clustering.

The invention discloses a multi-view clustering method based on anchor point-to-anchor graph structure collaborative regularization, and belongs to the technical field of multi-view clustering, and the method comprises the following steps: S1, building a model objective function; s2, optimizing an objective function; s3, analyzing algorithm complexity; according to the method, anchor point construction and anchor graph learning are combined into a unified framework: firstly, an implicit anchor point adjacency relation is constructed from an anchor graph, and a local smooth item is introduced to anchor points to guide anchor point learning; secondly, column sparsity is forced to be executed on the anchor map, each sample is encouraged to be only connected to several anchor points, and therefore redundant connection is reduced; and finally, Laplace rank constraint is applied to the anchor map so as to improve the clustering structure of the anchor map. The AGSCR-MVC uses the anchor point diagram as bridge representation and guides local structure learning and structure regularization of the anchor points at the same time, so that a two-stage collaborative modeling framework between an anchor point space and a diagram structure is realized, the clustering structure consistency between the anchor points and samples is enhanced, and the clustering performance is remarkably improved.

The invention discloses a structure perception granular ball information bottleneck method for multi-modal data clustering, which comprises the following steps: acquiring a multi-modal data set containing texts, pictures and voice segments, and performing granular ball division on the multi-modal data set to generate a granular ball set; for each particle ball, calculating a structure similarity ratio of the corresponding particle ball to the current cluster, and generating a structure sensing weight based on the structure similarity ratio; taking the pellets in the pellet set as a processing unit, and constructing an information bottleneck objective function which comprises an information retention item and an information compression item; and performing dynamic weighting on the information retention item by using the structure sensing weight, and iteratively optimizing the affiliation relation from the granular balls to the clustering cluster to maximize the weighted information bottleneck objective function until convergence, and outputting a final clustering result. According to the invention, a multi-modal clustering target with high robustness and high efficiency is realized.

Embodiments of the present application provide a kind of based on bill's graph clustering method, device, electronic equipment and readable storage medium, belong to computer technical field.The method includes: obtaining the multiple customer feature vectors of seed guest group based on bill;The cosine distance between any two customer feature vectors is calculated, and the initial adjacency matrix for representing customer similarity is determined according to each customer feature vector and the nearest K customer feature vector of corresponding cosine distance;The initial adjacency matrix is calculated by attention mechanism module, and a group of modified node features are obtained;A group of modified node features are carried out kmeans clustering calculation, and the cluster center of each clustering cluster is obtained, the similarity distribution result of cluster customer sample and cluster center in each clustering cluster is calculated;The final clustering result is calculated according to similarity distribution result.In this way, in bill scene, by preposition adding unsupervised graph clustering scheme, optimization clustering process, improve the clustering effect of final clustering result.