11 results about "Scatter matrix" patented technology

The invention provides a gas concentration time sequence distribution prediction method based on multivariable data driving. Based on the theoretical relationship between the mean square displacement and the mean scattering angle cosine of neutrons in an infinite homogeneous medium, the mean square displacement is directly counted through Monte Carlo simulation, the mean scattering angle cosine is reversely deduced, and a first-order scattering matrix meeting the mean square displacement conservation is constructed accordingly. In order to be suitable for a finite geometric model, a correction factor is further introduced to correct an average scattering angle cosine obtained by a traditional method. A verification result shows that the method remarkably reduces the calculation deviation of effective proliferation factors, improves the neutron flux distribution precision, and is particularly suitable for high-precision multi-group calculation of high-anisotropy fast spectrum reactor cores such as pebble-bed high-temperature gas cooled reactors and sodium cooled fast reactors. According to the method, the time sequence probability distribution prediction of the gas concentration can be realized, the prediction uncertainty is quantified, the risk indexes such as the over-limit probability are directly output, and the scientificity and the reliability of coal mine safety early warning are improved.

There is a desire for dimension compression of training data by linear discriminant analysis to be achieved by secure computation. An AI analysis model generation system according to the disclosed technology comprises a secure computation device and a user terminal. The secure computation device acquires data with n rows and m columns encrypted so as to be able to be securely computed, the data comprising n m-dimensional vectors to which class classifications are assigned. For each class, encrypted vectors belonging to the class are securely computed to obtain an encrypted sum vector and an encrypted sum-of-products matrix. The user terminal decrypts the encrypted sum vector and the encrypted sum-of-products matrix to obtain a plaintext sum vector and a plaintext sum-of-products matrix, uses the plaintext sum vector and the plaintext sum-of-products matrix to obtain a within-class scatter matrix and a between-class scatter matrix, and uses the within-class scatter matrix and the between-class scatter matrix to obtain a transformation matrix with m rows and d columns.

The application provides a high-dimensional small sample data dimension reduction method for industrial detection classification, acquires high-dimensional small sample data of industrial detection and converts the data into a data matrix; a local neighborhood structure is used to replace global statistics, a local inter-class scatter matrix and a local total scatter matrix are calculated, and based on this, a robust trace ratio discriminant local preserving projection model is constructed, so that the small sample problem is fundamentally avoided; an auxiliary variable is introduced to convert the model into a joint optimization problem about a projection matrix and the auxiliary variable, an iterative solution is obtained by using an alternating optimization strategy, convergence of the algorithm is ensured, and a final projection matrix is obtained; and test data is projected onto the final projection matrix to realize feature dimension reduction. The application improves the precision and robustness of industrial detection classification in a high-dimensional small sample scene.

The invention discloses a tertiary oil recovery layer series combination screening method based on an improved LDA method and a related device, and relates to the technical field of tertiary oil recovery, and the method comprises the steps: carrying out the preprocessing of original values of evaluation parameters in all feature data sets, and obtaining an evaluation matrix; calculating the evaluation matrix by using a global mean value method to obtain an overall mean value vector, and calculating to obtain intra-class and inter-class scatter matrixes based on the evaluation matrix and the overall mean value vector; calculating to obtain an optimal projection direction based on intra-class and inter-class scatter matrixes; based on the evaluation matrix and the optimal projection direction, calculating to obtain a comprehensive projection score corresponding to each tertiary oil recovery layer series combination; obtaining a feature weight corresponding to each evaluation parameter based on the feature data set; and on the basis of the comprehensive projection score, the evaluation matrix and the feature weight, calculating to obtain a comprehensive evaluation coefficient corresponding to each tertiary oil recovery layer series combination so as to screen out the layer series combination suitable for tertiary oil recovery of the three types of oil layers.

The invention discloses a linear multi-threshold classification method based on a Huffman tree, and the method comprises the steps: obtaining a correlation coefficient matrix according to multi-class data in a collected historical data set, and obtaining the average intensity of each class of data in a multi-feature dimension, so as to obtain the probability of each data class; according to the probability, coding is carried out through a Huffman tree, the coding length is inversely proportional to the probability, and classification containers are divided according to the coding length; and according to the divergence matrix of the data in the classification containers, in combination with the divergence matrix between the classification containers, obtaining a feature vector so as to obtain a regression mean value of the classification containers, and determining a classification threshold between the classification containers so as to perform classification by using the data collected in real time. According to the method, classification containers are dynamically divided based on Huffman tree coding, a multi-classification task is optimized into O (logK) subtasks, and in combination with divergence matrix driven feature vector optimization and dynamic threshold selection, the model complexity is reduced, and meanwhile, the recognition precision in a class imbalance scene is improved.

The invention belongs to the technical field of forest remote sensing, and discloses a P-band SAR polarization characteristic-based forest type identification method, which comprises the following steps of: solving a coherence matrix and a covariance matrix on the basis of a pre-processed polarization scattering matrix, and carrying out wavelength dependence correction on basic radar statistical characteristics for a forest three-layer vertical superposition scattering structure formed by a P band; extracting feature subsets of a scattering mechanism related to a forest vertical structure, integrating the corrected basic features and adding polarization phase difference and scattering mechanism combination coefficient feature dimensions to form a complete polarization feature set covering multiple types of features, and extracting various terrain parameters of a target forest region based on DEM data to obtain a complete polarization feature set covering multiple types of features; performing terrain sensitivity analysis on the preliminarily screened feature subset and dividing sensitivity grades, constructing an adaptive terrain correction model in combination with P-band under-forest surface scattering proportion characteristics, and introducing a radar wave polarization state correction term to complete feature radiation normalization correction.

The invention discloses a flame stability determination method and device. The method comprises the steps that physical field data related to the combustion process of target equipment is collected; extracting feature vectors of the temperature field data, the electric field data and the airflow field data to obtain a target temperature field feature vector, a target electric field feature vector and a target airflow field feature vector; combining the intra-group scatter matrix and the inter-group scatter matrix into a three-dimensional tensor comprising two channels through a dimension conversion layer of the first model, and compressing the high-dimensional feature data corresponding to the three-dimensional tensor into low-dimensional feature data; and determining a flame stability value of the target equipment according to the low-dimensional feature data through the multi-layer sensing layer of the first model. According to the scheme, multi-physical field information is fused, the first model is adopted for feature extraction, key data influencing flame stability can be effectively extracted, dimensionality reduction is carried out on high-dimensional complex data by optimizing the structure and the training method of the neural network, and the fidelity of key parameters in the dimensionality reduction process is ensured.

The application provides a well logging data anomaly detection method, which comprises the following steps: step 1, constructing a well logging data sample set; step 2, performing system initialization; step 3, performing feature extraction layer training; step 4, outputting classification layer training; and step 5, performing anomaly value identification. The well logging data anomaly detection method introduces a plurality of feature extraction layers, and introduces an intra-class scatter matrix when the feature extraction layers are constructed, so that abstract features with resolution can be obtained. When the output classification layer is trained, a risk weighting matrix is introduced, so that the robustness of the training result to noise is improved.

This invention discloses a kernelized inverse nearest neighbor discriminant analysis method, comprising the following steps: obtaining training image samples; mapping the input data to a high-dimensional space using a Gaussian kernel function; obtaining the representations of the intra-class and inter-class scatter matrices of the training image samples in the high-dimensional space using kernel tricks and the inverse nearest neighbor algorithm; learning the projection matrix that maximizes the inter-class scatter matrix and minimizes the intra-class scatter matrix through feature decomposition; extracting features from the training and test image samples using the projection matrix; and classifying the test samples using the nearest neighbor algorithm. This invention, for the first time, extends the inverse nearest neighbor linear discriminant analysis method to a high-dimensional space to solve the classification problem of nonlinear data. It proposes using a Gaussian kernel function for high-dimensional mapping, kernelizing the algorithm, and using kernel tricks for non-explicit mapping derivation. High-dimensional inverse nearest neighbors are established in the space after Gaussian function mapping to obtain the scatter matrix, wherein the kernel trick is used to realize the distance representation in the high-dimensional space.

The invention discloses a method for identifying adulteration of fish oil by combining Raman spectrum with multivariate statistical analysis. The method comprises the following steps: collecting the Raman spectrum of fish oil to be detected by using a Raman spectrometer; the collected Raman spectrum is subjected to preprocessing of denoising, baseline removal and normalization; dividing the spectral data into a training set and a test set according to a ratio of 7: 3; on the basis of traditional PCA, an intra-class scattering matrix and an inter-class scattering matrix are calculated, generalized Rayleigh quotients are calculated according to the intra-class scattering matrix and the inter-class scattering matrix, and sorting is carried out according to the generalized Rayleigh quotients; and training the linear discriminant analysis model by using data of the training set, and taking a result of the test set as a final classification result. Raman spectrum and machine learning are combined, solution consumption is low, sample preparation steps are simple, complex sample pretreatment is not needed, analysis time is short, and rapid detection and classification can be achieved. The analysis time is short, the classification accuracy is high, and the classification accuracy of the fish oil with the adulteration proportion of 5% reaches 95% or above.

The application provides a data dimension reduction method, a terminal device and a storage medium. The data dimension reduction method comprises the following steps: a coordinator obtains an encrypted first covariance matrix and first mean value data sent by an initiator, and obtains an encrypted second covariance matrix and second mean value data sent by a participant, the first mean value data is obtained by encrypting a first mean vector sum of all types of data in a first data set of the initiator, and the second mean value data is obtained by encrypting a second mean vector sum of all types of data in a second data set of the participant; an intra-class scatter matrix is determined according to the encrypted first covariance matrix and the encrypted second covariance matrix, an inter-class scatter matrix is determined according to the first mean value data and the second mean value data, a projection matrix is determined according to the intra-class scatter matrix and the inter-class scatter matrix, and the projection matrix is sent to the initiator and the participant, so that the data of the initiator and the participant is reduced in dimension while ensuring the security of the data of all parties.