Resistive material analysis method, apparatus, device, and product

Abstract

The application discloses a resistance material analysis method, device, equipment and product, relates to the technical field of data processing, and comprises the following steps: constructing an attribute feature matrix by using multi-source magnetic property data of a target resistance material; constructing an eigenvalue matrix based on eigenvalues between related attribute features in the attribute feature matrix; and performing clustering analysis on each eigenvalue in the eigenvalue matrix to obtain a magnetic property analysis result of the target resistance material. The application solves the technical problem of long analysis time and low accuracy of the magnetic resistance material, and achieves the technical effect of improving the magnetic property analysis efficiency and accuracy of the resistance material.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to resistance material analysis methods, devices, equipment and products. Background Art

[0002] As an important functional material, magnetoresistance materials have broad application prospects in fields such as sensors, memory, and energy transmission. With the rapid development of devices, extremely large magnetoresistance materials have also emerged. Due to the complex microscopic mechanism of extremely large magnetoresistance materials, the relationship between their resistivity change and magnetic field intensity is nonlinear, which poses challenges to the analysis and application of materials. The analysis of magnetoresistance materials in related technologies often relies on the extremely large magnetoresistance effect test of the resistive material and single-material simulation. Although it can provide information on material properties to a certain extent, it is often time-consuming and has low accuracy. Summary of the Invention

[0003] The present application provides a method, apparatus, equipment and product for analyzing resistance materials to at least solve the problem that the analysis of magnetoresistance materials in related technologies often relies on the maximum magnetoresistance effect test and single material simulation of the resistance materials. Although these methods can provide information on the material properties to a certain extent, they are often time-consuming and have low accuracy.

[0004] This application provides a resistance material analysis method, comprising:

[0005] Utilize multi-source magnetic property data of target resistive materials to construct a property feature matrix;

[0006] Based on the eigenvalues ​​between the related attribute features in the attribute feature matrix, an eigenvalue matrix is ​​constructed;

[0007] Cluster analysis is performed on each eigenvalue in the eigenvalue matrix to obtain the magnetic property analysis results of the target type of resistor material.

[0008] In an optional embodiment, a property feature matrix is ​​constructed using multi-source magnetic property data of the target resistive material, including:

[0009] Obtain magnetic property test data and magnetic property theoretical data of the target type resistor material, where the multi-source magnetic property data includes magnetic property test data and magnetic property theoretical data;

[0010] Process the magnetic property test data and magnetic property theoretical data to obtain target magnetic property data;

[0011] Construct an attribute feature matrix based on the target magnetic attribute data.

[0012] In an optional embodiment, the magnetic property test data and the magnetic property theoretical data are processed to obtain target magnetic property data, including:

[0013] Obtaining missing data and abnormal data in magnetic property test data and magnetic property theoretical data after data alignment;

[0014] Perform data cleaning on magnetic property test data and magnetic property theoretical data based on missing data and abnormal data to obtain magnetic property cleaned data;

[0015] The magnetic property cleaning data is subjected to data standardization processing to obtain target magnetic property data.

[0016] In an optional embodiment, constructing an eigenvalue matrix based on eigenvalues ​​between related attribute features in the attribute feature matrix includes:

[0017] Acquire multiple attribute features that are attribute-correlated and / or attribute-negatively correlated in the attribute feature matrix;

[0018] Constructing a covariance matrix between multiple attribute features with correlated and / or negatively correlated attributes;

[0019] Construct an eigenvalue matrix based on the eigendecomposition values ​​of the covariance matrix.

[0020] In an optional embodiment, constructing an eigenvalue matrix based on the eigendecomposition values ​​of the covariance matrix includes:

[0021] Perform eigenvalue decomposition on the covariance matrix to obtain the eigendecomposition value;

[0022] An eigenvalue matrix is ​​constructed based on the eigenvectors corresponding to the target eigendecomposition values ​​in the eigendecomposition values.

[0023] In an optional implementation, constructing an eigenvalue matrix based on eigenvectors corresponding to target eigendecomposition values ​​in the eigendecomposition values ​​includes:

[0024] Calculate the cumulative variance contribution rate of each eigendecomposition value;

[0025] Construct a projection matrix based on the eigenvectors corresponding to the target eigendecomposition values ​​whose cumulative variance contribution is greater than the contribution threshold;

[0026] Based on the attribute feature matrix and the projection matrix, the eigenvalue matrix is ​​constructed.

[0027] In an optional implementation, constructing an eigenvalue matrix based on eigenvectors corresponding to target eigendecomposition values ​​in the eigendecomposition values ​​includes:

[0028] Draw a characteristic curve of the eigendecomposition value as it changes with the number of principal components;

[0029] Construct a projection matrix based on the eigenvectors corresponding to the target eigendecomposition values ​​whose slopes in the characteristic curve are greater than the slope threshold;

[0030] Based on the attribute feature matrix and the projection matrix, the eigenvalue matrix is ​​constructed.

[0031] In an optional embodiment, cluster analysis is performed on each eigenvalue in the eigenvalue matrix to obtain magnetic property analysis results of the target type resistor material, including:

[0032] Performing cluster analysis on each eigenvalue in the eigenvalue matrix to obtain a clustering result of magnetic properties of each resistor material in the target class of resistor materials and / or a clustering result of magnetic correlations between target class materials;

[0033] Based on the magnetic property clustering results and / or the magnetic correlation clustering results, the magnetic property analysis results of the target type resistor material are obtained.

[0034] In an optional embodiment, cluster analysis is performed on each eigenvalue in the eigenvalue matrix to obtain a magnetic property analysis result of the target type resistor material, further comprising:

[0035] The target magnetic property clustering model is used to perform cluster analysis on the eigenvalues ​​in the eigenvalue matrix to obtain the magnetic property analysis results of the target type resistor material.

[0036] In an optional embodiment, obtaining a target magnetic property clustering model includes:

[0037] Obtaining attribute feature training samples of target resistor materials;

[0038] Input the attribute feature training samples into the Gaussian mixture model to obtain the initial posterior probability that each attribute feature training sample belongs to the corresponding Gaussian component. The model parameters of the Gaussian mixture model include the initial mean vector, the initial covariance matrix and the initial feature weight;

[0039] The model parameters of the Gaussian mixture model are updated based on the initial posterior probability to obtain a target magnetic property clustering model.

[0040] In an optional embodiment, updating the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain a target magnetic property clustering model includes:

[0041] Calculate the expected maximization value of each attribute feature training sample, which is used to characterize the contribution of each attribute feature training sample to each Gaussian component in the Gaussian mixture model;

[0042] Adjust the initial mean vector, initial covariance matrix and initial feature weight according to the expected maximization value to obtain the updated mean vector, covariance matrix and feature weight;

[0043] The step of calculating the expectation maximization value of each attribute feature training sample is iteratively performed until the model parameters of the Gaussian mixture model converge, thereby obtaining a target magnetic attribute clustering model.

[0044] In an optional embodiment, updating the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain a target magnetic property clustering model includes:

[0045] Compare the initial posterior probability with the target posterior probability to obtain a comparison result;

[0046] If the comparison result indicates that the initial posterior probability is less than the target posterior probability, the model parameters of the Gaussian mixture model are updated to the target model parameters based on the difference between the initial posterior probability and the target posterior probability, and the degree of correlation between each attribute feature training sample and the maximum magnetoresistance effect, to obtain the target magnetic attribute clustering model.

[0047] The present application also provides a resistance material analysis device, comprising:

[0048] A data acquisition module is used to construct a property feature matrix using multi-source magnetic property data of target resistor materials;

[0049] A feature acquisition module is used to construct an eigenvalue matrix based on the eigenvalues ​​between related attribute features in the attribute feature matrix;

[0050] The material analysis module is used to perform cluster analysis on each eigenvalue in the eigenvalue matrix to obtain the magnetic property analysis results of the target type of resistor material.

[0051] The present application also provides a computer device, comprising: a memory for storing a computer program; and a processor for implementing the steps of any of the above-mentioned resistance material analysis methods when executing the computer program.

[0052] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned resistance material analysis methods are implemented.

[0053] The present application also provides a computer program product, including a computer program, which implements the steps of any of the above-mentioned resistance material analysis methods when executed by a processor.

[0054] Through the resistance material analysis method of the present application, by utilizing the multi-source magnetic property data of the target type of resistance material, and constructing the property feature matrix and the eigenvalue matrix, and then performing cluster analysis on the eigenvalues ​​in the eigenvalue matrix, it is possible to comprehensively and quickly obtain the magnetic property analysis results of the target type of resistance material, which not only improves the analysis efficiency, but also can more accurately capture the complex behavior of the material under different conditions, meeting the application needs of the analysis of the maximum magnetoresistance effect of the resistance material. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0056] Figure 1 A schematic diagram of a process for analyzing a resistor material according to an embodiment of the present application;

[0057] FIG2( a ) is a schematic diagram of the energy bands of the resistor material provided in an embodiment of the present application;

[0058] FIG2( b ) is a schematic diagram of the Fermi surface of the resistor material provided in an embodiment of the present application;

[0059] Figure 3 A schematic diagram of the cyclotron trajectory of the Fermi surface of the carrier provided in an embodiment of the present application;

[0060] Figure 4 A schematic diagram of outlier verification based on Hall conductance results provided in an embodiment of the present application;

[0061] Figure 5 A schematic diagram of drawing characteristic curves provided in an embodiment of the present application;

[0062] Figure 6 Another schematic diagram of a process for analyzing a resistor material according to an embodiment of the present application;

[0063] Figure 7 A structural block diagram of a resistance material analysis device provided in an embodiment of the present application;

[0064] Figure 8 A schematic diagram of the structure of a computer device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0065] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0066] It should be noted that, in the description of this application, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. The terms "first," "second," etc., in this application are used to distinguish similar objects, and are not used to describe a particular order or sequence.

[0067] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0068] In this embodiment, a method for analyzing resistance materials is provided. Figure 1 FIG. 1 is a flow chart of a resistance material analysis method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0069] Step 110: construct an attribute feature matrix using multi-source magnetic attribute data of the target type resistor material.

[0070] The target resistive material can be any material exhibiting magnetoresistance, including but not limited to metal alloys, oxides, and multilayer film structures. Multi-source magnetic property data can include but is not limited to magnetic property test data such as magnetoresistance, magnetization, coercivity, carrier concentration, mobility, crystal structure, symmetry, and temperature dependence, as well as theoretical magnetic property data calculated based on quantum mechanics and solid-state physics. The construction of a property feature matrix can organize complex multi-source magnetic property data into a structured form, facilitating subsequent analysis and processing.

[0071] In some optional embodiments, when constructing a property feature matrix using multi-source magnetic property data of a target type of resistor material, the magnetic property test data and magnetic property theoretical data of the target type of resistor material can be first obtained; the magnetic property test data and magnetic property theoretical data can be processed to obtain target magnetic property data; and the property feature matrix can be constructed based on the target magnetic property data.

[0072] Specifically, to obtain the magnetic property test data and theoretical magnetic property data for a target resistor material, the magnetic property information can be retrieved by querying a material database. This database can be a public database or a self-built internal database. The database stores a large amount of magnetic property test data and theoretical data for resistor materials, which has been rigorously screened and verified to ensure accuracy and reliability. By querying this database, the magnetic property test data and theoretical magnetic property data for the target resistor material can be quickly obtained, facilitating subsequent data processing and analysis.

[0073] Furthermore, if the required magnetic property data of the target class resistor material does not exist in the database, the magnetic property data of the target class resistor material can also be obtained by testing the maximum magnetoresistance effect of the resistor material or by theoretical calculation. The maximum magnetoresistance effect test of the resistor material can be performed by various magnetic measuring instruments, such as vibrating sample magnetometers, superconducting quantum interference devices, etc., to test the magnetic properties of the target class resistor material and obtain magnetic property test data. Theoretical calculation can be performed by computer simulation based on density functional theory, Monte Carlo simulation and other methods to predict and calculate the magnetic properties of the target class resistor material and obtain magnetic property theoretical data. After obtaining the magnetic property test data and magnetic property theoretical data, these data need to be processed to obtain the target magnetic property data. Data processing may include but is not limited to steps such as data alignment, data cleaning, and data standardization to ensure the accuracy and consistency of the data. After obtaining the target magnetic property data, an attribute feature matrix can be constructed based on these data. The construction of the attribute feature matrix can be to arrange and combine the target magnetic property data according to certain rules to form a two-dimensional matrix form. Each row of the matrix can represent the magnetic property data of a target class resistor material, and each column can represent different magnetic property characteristics. By constructing the attribute feature matrix, complex multi-source magnetic attribute data can be converted into a structured form, which facilitates the subsequent eigenvalue matrix construction and cluster analysis.

[0074] In some optional embodiments, when data processing is performed on the magnetic property test data and the magnetic property theoretical data to obtain the target magnetic property data, the missing data and abnormal data in the magnetic property test data and the magnetic property theoretical data after data alignment can be first obtained; the magnetic property test data and the magnetic property theoretical data are cleaned based on the missing data and the abnormal data to obtain magnetic property cleaned data; and the magnetic property cleaned data are data standardized to obtain the target magnetic property data.

[0075] Specifically, the magnetic property test data and the magnetic property theoretical data can be aligned first to ensure the accuracy and consistency of the data. The data alignment process can be implemented based on timestamps, sample numbers or other unique identifiers, so that the test data and theoretical data can correspond one to one. Then identify the unfilled or blank entries in the data set, which may be caused by equipment failure, operational errors or other reasons in the data recording process. At the same time, detect outliers in the data, which may appear as extreme values, inconsistent with expected data patterns or contrary to known physical laws. For missing data, interpolation, regression prediction, estimation based on neighboring data, database search and other methods can be used to fill in the missing data based on the attribute characteristics of the actual data to ensure the integrity of the data. As for abnormal data, they are corrected or eliminated according to the specific situation to avoid adverse effects on subsequent analysis.

[0076] For abnormal data, complex attribute engineering can be initiated to further clean the data through expert interaction. For example, in theory, the band structure can be used to deduce any property characteristics of the material. However, in reality, even if the band characteristics of the resistive material are obtained through calculation, it is not possible to accurately determine what properties each band exhibits. For example, obtaining the carrier symbol characteristics is extremely complex, and in Figure 2 (a) Directional band properties and The directional energy band performance is difficult to judge by its slope. The Fermi surfaces corresponding to Figure 2 (b) are the α position and Position, its properties can be determined by calculating its rotation trajectory, such as Figure 3 , representing the trajectories of the two orbits respectively. The orbit moves counterclockwise around the magnetic field, so it can be judged that the nature of this orbit is electron type; the orbit moves clockwise around the magnetic field, and its orbit nature is hole type.

[0077] In addition, data cross-validation can be used to obtain abnormal data in the aligned magnetic property test data and magnetic property theoretical data to ensure the accuracy of the numerical values ​​and physical properties. Whether the numerical simulation results are consistent with the test data can be judged by the sign of the Hall resistance. The expression of Hall conductivity is:

[0078]

[0079] in, represents the conductivity of the energy band n under the action of the magnetic field, e represents the number of electrons, represents the relaxation time of energy band n, represents the energy of band n corresponding to the wave vector k in reciprocal space, It represents the velocity of the electron wave packet of the carrier under the action of the magnetic field at the reciprocal space wave vector k, represents the average velocity of the electron wave packet in the mesoscopic system under the action of the magnetic field, represents the total conductance contributed by all states on the calculated Fermi surface to the transport.

[0080]

[0081] in, represents Planck's constant, represents the velocity of the electron wave packet of the mesoscopic system under the action of the magnetic field at the reciprocal space wave vector k, R represents the position variable, represents the phase factor, and represents the transformation matrix under energy representation, Represents the Hamiltonian of the mesoscopic system in the reciprocal space for any wave vector. The Hall conductance results obtained after calculation are as follows Figure 4 As shown in Figure 1, where elec is the abbreviation for electron carriers, hole is the abbreviation for hole carriers, and xx and xy represent the components. A negative Hall conductance value indicates electron carriers (electron type), while a negative value indicates hole carriers (hole type). Complex energy bands may experience type transitions, i.e., sign shifts. The sign of the effective mass of the extremal orbital can be used to determine whether the band is electron or hole type, or even more complex multi-band properties.

[0082] Furthermore, when normalizing the cleaned magnetic property data to obtain the target magnetic property data, a data normalization algorithm can be used to eliminate outliers and noise in the data, ensuring data accuracy and reliability. Simultaneously, data with different dimensions are converted to the same dimension, facilitating subsequent data analysis and processing. Data normalization effectively improves the processing efficiency and accuracy of magnetic property data, providing strong support for subsequent data analysis and determination of resistor material properties.

[0083] For example, temperature or magnetic field dependence, or the coefficients of a polynomial fit, can be used as features. Categorical variables are then encoded, such as lattice structure, cubic structure = 1000, hexagonal structure = 0100, etc. Redundant categories such as complex monoclinic and triclinic structures can be combined into low-symmetry structures = 0001. The attribute feature matrix is ​​constructed based on the target magnetic attribute data as follows:

[0084]

[0085] Step 120: construct an eigenvalue matrix based on the eigenvalues ​​between related attribute features in the attribute feature matrix.

[0086] The relevant attribute features in the attribute feature matrix include magnetoresistance, coercivity, carrier concentration, mobility, and so on. The eigenvalue matrix can be constructed by calculating the correlation or similarity between the individual attribute features in the attribute feature matrix, resulting in a two-dimensional matrix. Each row and column of the matrix represents a different attribute feature, and the elements in the matrix represent the eigenvalues ​​between the corresponding attribute features. Eigenvalues ​​can be correlation coefficients, distance metrics, similarity indices, and other indicators that describe the degree of association or difference between attribute features. By constructing the eigenvalue matrix, the relationships between attribute features can be further explored, providing a basis for subsequent cluster analysis.

[0087] In some optional embodiments, when constructing an eigenvalue matrix based on the eigenvalues ​​between related attribute features in the attribute feature matrix, multiple attribute features that are attribute-related and / or attribute-negatively correlated in the attribute feature matrix can be obtained; a covariance matrix between multiple attribute features that are attribute-related and / or attribute-negatively correlated is constructed; and an eigenvalue matrix is ​​constructed based on the eigendecomposition values ​​of the covariance matrix.

[0088] During this process, the calculation of the covariance matrix can reflect the strength and direction of the linear relationship between attribute features. Positive attribute correlation means that the two attribute features change in the same trend, while negative attribute correlation indicates that they change in opposite trends. Through eigendecomposition, the covariance matrix is ​​converted into eigenvalues ​​and eigenvectors, which can reveal the main direction and degree of variation in the data. Using these eigendecomposition values, an eigenvalue matrix is ​​constructed, which more intuitively demonstrates the correlation patterns between attribute features. This not only improves the accuracy of data analysis, but also enhances the model's ability to predict resistor material properties, while also providing strong support for subsequent data analysis and model training.

[0089] As an example, for an n-dimensional attribute feature matrix X (with m samples), the covariance matrix C is calculated as C=1 / (m-1)X T X. Example: If “mobility” is negatively correlated with “effective mass”, the corresponding element in the covariance matrix will be negative.

[0090] In some optional embodiments, when constructing an eigenvalue matrix based on the eigendecomposition values ​​of the covariance matrix, the covariance matrix can be subjected to eigenvalue decomposition to obtain eigendecomposition values; and the eigenvalue matrix can be constructed based on the eigenvectors corresponding to the target eigendecomposition values ​​in the eigendecomposition values. The target eigendecomposition values ​​can be selected based on actual needs, for example, the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to retain the main variation information in the data, while reducing the dimensionality of the data and improving the efficiency of subsequent model training and data analysis.

[0091] As an example, based on the eigenvectors corresponding to the target eigenvalues ​​in the eigenvalues, before constructing the eigenvalue matrix, the covariance matrix C can be first eigenvalue decomposition to obtain the eigenvalues ​​λ1,λ2…λn and the corresponding eigenvectors v1,v2…vn. The eigenvalue represents the variance of the data in the direction of the corresponding eigenvector. The larger the variance, the more data information is contained in that direction. The eigenvalues ​​are sorted from large to small, and the corresponding eigenvectors are also arranged in the same order. The sorted eigenvalues ​​and eigenvectors can be expressed as λ(1)≥λ(2)≥…≥λ(n) and v(1),v(2)…v(n). For example, the first principal component (PC1) explains 60% of the variance, which may correspond to "carrier transport capacity" (determined by mobility and effective mass). The second principal component (PC2) explains 25% of the variance, which may correspond to "material structure symmetry."

[0092] In some optional embodiments, when constructing an eigenvalue matrix based on the eigenvectors corresponding to the target class eigendecomposition values ​​in the eigendecomposition values, the cumulative variance contribution rate of each eigendecomposition value can be calculated first; based on the eigenvectors corresponding to the target eigendecomposition values ​​whose cumulative variance contribution is greater than the contribution threshold, a projection matrix can be constructed; and based on the attribute feature matrix and the projection matrix, an eigenvalue matrix can be constructed.

[0093] The cumulative variance contribution rate reflects the proportion of the total information contained in the first k principal components to the total information. Using the contribution threshold, we can filter out eigenvalues ​​that contribute significantly to the variance, thereby retaining the primary information and removing noise or secondary information. The projection matrix is ​​constructed based on the selected eigenvectors. It projects the original data into a low-dimensional space while preserving the data's primary features as much as possible. Finally, by combining the attribute feature matrix and the projection matrix, we can construct the eigenvalue matrix. This matrix not only has a lower dimensionality but also contains the primary information in the original data, facilitating subsequent analysis and processing.

[0094] As an example, the formula for calculating the cumulative variance contribution of the first m eigendecomposition values ​​is S= Typically, the projection matrix is ​​constructed by selecting the eigenvectors corresponding to the eigendecomposition values ​​when the cumulative variance contribution reaches 80%-95%. This effectively reduces the data dimension while minimizing information loss, improving the efficiency of subsequent processing and analysis. The projection matrix construction process essentially maps the original high-dimensional data into a low-dimensional space generated by the selected eigenvectors. In this low-dimensional space, the data's primary features are preserved, while noise or secondary information is removed or attenuated.

[0095] After obtaining the projection matrix, multiply it by the original attribute feature matrix to obtain the eigenvalue matrix. Each row and column of the eigenvalue matrix represents the projection of different attribute features in low-dimensional space, and the elements in the matrix reflect the degree of correlation or difference between these projections. The eigenvalue matrix allows for more intuitive analysis of the relationships between attribute features, providing strong support for subsequent data analysis and model training.

[0096] Furthermore, when constructing the eigenvalue matrix, weight coefficients can be introduced to adjust the contribution of different attribute features to the eigenvalue matrix. Weight coefficients can be determined based on practical needs, for example, based on factors such as the importance, relevance, or physical significance of the attribute features. By introducing weight coefficients, the eigenvalue matrix can be made more consistent with practical application requirements, improving the accuracy and reliability of data analysis.

[0097] In summary, by constructing an eigenvalue matrix, we can further explore the relationships between attribute features, providing strong support for subsequent data analysis and the determination of resistor material properties. Furthermore, through methods such as eigenvalue decomposition, projection matrix construction, and the introduction of weight coefficients, we can optimize the eigenvalue matrix construction process and improve the efficiency and accuracy of data analysis.

[0098] In some optional embodiments, when constructing an eigenvalue matrix based on the eigenvectors corresponding to the target eigendecomposition values ​​in the eigendecomposition values, a characteristic curve in which the eigendecomposition values ​​change with the number of principal components can also be plotted; a projection matrix can be constructed based on the eigenvectors corresponding to the target eigendecomposition values ​​whose slopes in the characteristic curves are greater than the slope threshold; and an eigenvalue matrix can be constructed based on the attribute feature matrix and the projection matrix.

[0099] When drawing the characteristic curve, you can use visualization tools such as Matlab and Python's Matplotlib to graphically display the changes in the characteristic decomposition value with the number of principal components, such as Figure 5 As shown in the figure. By observing the changes in the slope of the characteristic curve, it is possible to intuitively identify which eigenvalues ​​have a significant change trend as the number of principal components increases. When the slope of the characteristic curve is greater than the slope threshold, the corresponding eigenvalue is considered to have a significant impact on the construction of the eigenvalue matrix. Therefore, the eigenvectors corresponding to these eigenvalues ​​are selected to construct the projection matrix. The construction of the projection matrix helps to further extract key information, reduce data dimensions, and improve the efficiency of data analysis. Ultimately, based on the screened attribute feature matrix and projection matrix, a more accurate and efficient eigenvalue matrix is ​​constructed, providing strong support for in-depth analysis of resistor material properties.

[0100] As an example, select the first m eigenvectors v(1), v(2)…v(m) to form the projection matrix P=[v(1), v(2)…v(m)]. Multiply the normalized attribute feature matrix X by the projection matrix P to obtain the reduced-dimensional feature matrix Y=XP.

[0101] Step 130 : performing cluster analysis on each eigenvalue in the eigenvalue matrix to obtain a magnetic property analysis result of the target type resistor material.

[0102] Cluster analysis can employ a variety of algorithms, such as K-means clustering, hierarchical clustering, and density clustering. The specific choice depends on the characteristics of the data and the analysis requirements. During the clustering process, the algorithm groups similar properties into the same category based on the values ​​of the elements in the eigenvalue matrix, thereby revealing the inherent laws and patterns of the magnetic properties of the resistive material. Clustering results can be displayed graphically, such as cluster dendrograms, scatter plots, boxplots, and t-distributed random neighborhood embedding, for intuitive understanding and analysis.

[0103] In some optional embodiments, when clustering analysis is performed on each eigenvalue in the eigenvalue matrix to obtain the magnetic property analysis results of the target type resistor material, cluster analysis can be performed on each eigenvalue in the eigenvalue matrix to obtain the magnetic property clustering results of each resistor material in the target type resistor material and / or the magnetic correlation clustering results between the target type materials; based on the magnetic property clustering results and / or the magnetic correlation clustering results, the magnetic property analysis results of the target type resistor material are obtained.

[0104] Specifically, cluster analysis is performed on the eigenvalues ​​in the eigenvalue matrix. When clustering the magnetic properties of each resistor within a target class and / or clustering the magnetic correlations between materials within the target class, a density-based DBSCAN algorithm or a grid-based CLIQUE algorithm can be used to quickly identify similarities and correlations between eigenvalues. High-performance graphics processing units (GPUs) can also be configured to accelerate the clustering calculation process, ensuring accurate clustering results within a reasonable timeframe. Furthermore, the clustering results can be presented intuitively, such as through dynamic interactive scatter plots or heat maps, to help users gain a deeper understanding of the magnetic properties of the target class of resistors and the relationships between them. Because magnetic property clustering reveals similarities and differences in magnetic properties within the same class of materials, it helps identify groups of materials with similar magnetic properties. Furthermore, magnetic correlation clustering between target materials reveals the degree of correlation and mutual influence between the magnetic properties of different materials, providing important insights into material interactions and performance regulation. By comprehensively analyzing these clustering results, we can gain a deeper understanding of the magnetic properties of the target resistor materials and provide a scientific basis for material screening, optimization and design.

[0105] Furthermore, based on cluster analysis, the relationship between magnetic properties and material composition can be analyzed by combining the material's microstructure and chemical composition, revealing the origin and regulation mechanism of magnetism. Furthermore, the magnetic property analysis results can be combined with other physical properties of the material, such as electrical and thermal properties, to conduct multi-dimensional performance evaluation and comparison, providing guidance for optimizing the material's comprehensive performance.

[0106] Furthermore, cluster analysis results can provide strong support for data mining and machine learning model training for resistor materials. By inputting clustering results as features into machine learning models, the model's prediction accuracy and generalization capabilities for resistor material properties can be improved. Furthermore, cluster analysis results can serve as an important step in data preprocessing, helping to reduce data redundancy and noise and improve the efficiency and accuracy of data analysis.

[0107] The resistor material analysis method provided by the present invention utilizes multi-source magnetic property data of the target resistor material, constructs a property feature matrix and an eigenvalue matrix, and then performs cluster analysis on the eigenvalues ​​in the eigenvalue matrix. This method can comprehensively and quickly obtain magnetic property analysis results of the target resistor material. This not only improves analysis efficiency but also more accurately captures the complex behavior of the material under different conditions, meeting the application requirements for extremely large magnetoresistance (XMR) analysis of resistor materials.

[0108] Figure 6 FIG. 2 is a flow chart showing another embodiment of the resistance material analysis method of the present invention. Figure 6 As shown, the method includes the following steps:

[0109] Step 610: construct an attribute feature matrix using multi-source magnetic attribute data of the target type resistor material.

[0110] For details, please see Figure 1 Step 110 of the illustrated embodiment will not be described in detail here.

[0111] Step 620: construct an eigenvalue matrix based on the eigenvalues ​​between the related attribute features in the attribute feature matrix.

[0112] For details, please see Figure 1 Step 120 of the illustrated embodiment will not be described in detail here.

[0113] Step 630 : Perform cluster analysis on each eigenvalue in the eigenvalue matrix using the target magnetic property clustering model to obtain a magnetic property analysis result of the target type resistor material.

[0114] Specifically, the above step 630 includes:

[0115] Step 6301: Obtain attribute feature training samples of target class resistor materials;

[0116] Step 6302: Input the attribute feature training samples into the Gaussian mixture model to obtain the initial posterior probability that each attribute feature training sample belongs to the corresponding Gaussian component. The model parameters of the Gaussian mixture model include the initial mean vector, the initial covariance matrix, and the initial feature weights.

[0117] Step 6303: Update the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain a target magnetic property clustering model.

[0118] Specifically, the attribute characteristic training samples of the target type resistor material can obtain attribute characteristic data related to the target type resistor material from public databases or literature. These data may come from the maximum magnetoresistance effect test of the resistor material, simulation calculation or other research methods. In addition, it is also possible to cooperate with enterprises or research institutions to obtain their unique test data and research results. The acquired attribute characteristic training samples cover the main attribute characteristics of the target type resistor material, such as magnetic permeability, resistivity, saturation magnetization, etc., as well as the changes of these attribute characteristics under different conditions. At the same time, the training samples are representative samples that can reflect the overall characteristics and change trends of the target type resistor material. After obtaining enough attribute characteristic training samples, these samples can be input into the Gaussian mixture model for training. Among them, the initial mean vector, initial covariance matrix and initial feature weights can be random values, initial values ​​set according to prior knowledge, or obtained by clustering the attribute characteristic training samples through the k-means algorithm.

[0119] The Gaussian mixture model (GMM) is a clustering model based on probabilistic statistics. It assumes that all data points are generated by several Gaussian distributions, with each Gaussian distribution representing a cluster. During training, the GMM continuously adjusts the mean vector, covariance matrix, and feature weights of each Gaussian distribution to maximize the posterior probability of the data point under the corresponding Gaussian distribution. Initially, a random mean vector, covariance matrix, and feature weight are set for each Gaussian component. Based on these initial parameters, the initial posterior probability of each attribute feature training sample belonging to the corresponding Gaussian component is calculated. Next, using the expectation-maximization algorithm (EM) or other optimization algorithm, the GMM parameters are iteratively updated until convergence or the preset number of iterations is reached. In each iteration, the mean vector, covariance matrix, and feature weights of the Gaussian distributions are recalculated based on the current posterior probability and the attribute feature training samples. Through continuous iterative optimization, a more accurate target magnetic attribute clustering model can be obtained. This target magnetic property clustering model can group similar property characteristics into the same category based on the eigenvalues ​​in the eigenvalue matrix, thereby revealing the inherent laws and patterns of the magnetic properties of resistor materials. Using the trained target magnetic property clustering model, the eigenvalue matrix of a new target class of resistor materials can be clustered and analyzed to quickly obtain their magnetic property analysis results. This not only improves analysis efficiency but also more accurately captures the complex behavior of materials under different conditions, providing a scientific basis for material screening, optimization, and design.

[0120] In some optional embodiments, when the model parameters of the Gaussian mixture model are updated based on the initial posterior probability to obtain the target magnetic attribute clustering model, the expectation maximization value of each attribute feature training sample can be calculated, and the expectation maximization value is used to characterize the contribution of each attribute feature training sample to each Gaussian component in the Gaussian mixture model; the initial mean vector, initial covariance matrix and initial feature weight are adjusted according to the expectation maximization value to obtain the updated mean vector, covariance matrix and feature weight; the step of calculating the expectation maximization value of each attribute feature training sample is iteratively performed until the model parameters of the Gaussian mixture model converge to obtain the target magnetic attribute clustering model.

[0121] During the iteration process, the model's convergence can be monitored, such as by observing the trend of the log-likelihood function value. When the log-likelihood function value stabilizes, the model is considered converged. Furthermore, a maximum number of iterations can be set to prevent the model from continuing to run if it fails to converge. Once a converged Gaussian mixture model is obtained, it can be used as the target magnetic property clustering model to perform cluster analysis on the eigenvalue matrix of a new target resistor material. The results of the cluster analysis will reveal the magnetic property characteristics and inherent laws of the target resistor material, providing strong support for further research and application of the material.

[0122] As an example, based on the Gaussian mixture model estimation K The mean of a Gaussian distribution ( μ ), covariance ( Σ ) and weights ( π Before training the Gaussian mixture model, initialize the three core parameters of the Gaussian mixture model:

[0123] The initial mean vector of each Gaussian component ( μ 1, μ 2 …μ M )

[0124] The initial covariance matrix of each Gaussian component ( Σ 1, Σ 2… Σ M )

[0125] The initial feature weights of each Gaussian component ( π 1, π 2… π M ,and Σ π_i =1)

[0126] Initialization: Select K samples from the data as the initial mean, initialize the covariance matrix to the identity matrix, and initialize the weights.

[0127] Adjust weights based on physical mechanism association: In the target class of resistive materials, namely the extremely large magnetoresistance materials, different properties have different influences on the extremely large magnetoresistance effect. For example, properties such as carrier compensation mechanism, mobility, and effective mass play a key role in the effect. Therefore, for the Gaussian components closely related to these key mechanisms, their weights are π i They should not be simply initialized to equal values, but can be given greater weight based on prior knowledge or preliminary experimental results. For example, if previous studies have shown that carrier compensation is extremely important in the maximum magnetoresistance effect, then the Gaussian component π related to carrier compensation can be i It can be increased appropriately.

[0128] When calculating the initial posterior probability that each attribute feature training sample belongs to the corresponding Gaussian component, for each attribute feature training sample x j , the model for calculating the probability that it belongs to the i-th Gaussian component is:

[0129]

[0130] in, is the probability density function of the Gaussian distribution.

[0131] Since there are complex correlations between multiple properties of target resistive materials, the covariance matrix The physical correlation between the band structure and the carrier mobility can be measured, and when calculating the covariance matrix, the covariance between the two properties can be given additional weight to better reflect their synergistic effect in the giant magnetoresistance effect. Consider the attribute range constraint: some properties of the target resistive material have specific value ranges. For example, the carrier concentration will exhibit significant giant magnetoresistance effect only within a certain range. Therefore, the probability density function can be modified, and when the sample attribute value exceeds the reasonable range, the corresponding probability density value is reduced. For example, if the carrier concentration exceeds the normal range, it can be adjusted by multiplying it by a decay function . .

[0132] Introduce a giant magnetoresistance effect indicator: introduce an indicator that is directly related to the giant magnetoresistance effect of the sample itself , such as the magnetoresistance change rate of the material. Then the posterior probability model can be adjusted as follows: , which can make the model pay more attention to sample data that is strongly related to the giant magnetoresistance effect and highlight the role of the giant magnetoresistance effect in clustering. Where represents the "contribution degree" of sample j to the i-th component, and the value range is [0, 1].

[0133] Further, update the model parameters of the Gaussian mixture model based on the initial posterior probability to maximize the log-likelihood function. Including:

[0134] The initial feature weight update model is:

[0135]

[0136] In addition, the weight can also be dynamically adjusted according to the degree of association between the properties of the target resistive material and the giant magnetoresistance effect. That is, an adjustment factor α i , α i is introduced. The value of the adjustment factor is related to the contribution of the physical mechanism represented by the component to the giant magnetoresistance effect. Then the corresponding new initial feature weight update model is , and at each iteration, the influence of the attribute on the giant magnetoresistance effect in the new data is updated in real time α i , that is, for the same material attribute, the adjustment factor changes at each iteration.

[0137] The new model for updating the initial mean vector is:

[0138]

[0139] The update model of the initial covariance matrix (taking full covariance as an example) is:

[0140]

[0141] Furthermore, when selecting the covariance matrix, you can choose an independent covariance matrix for each component, or you can choose the same covariance matrix or a diagonal covariance matrix shared by all components, or you can choose an isotropic covariance matrix.

[0142] In some optional embodiments, when the model parameters of the Gaussian mixture model are updated based on the initial posterior probability to obtain the target magnetic property clustering model, the initial posterior probability can also be compared with the target posterior probability to obtain a comparison result; if the comparison result indicates that the initial posterior probability is less than the target posterior probability, the model parameters of the Gaussian mixture model are updated to the target model parameters based on the difference between the initial posterior probability and the target posterior probability, and the degree of correlation between each attribute feature training sample and the maximum magnetoresistance effect to obtain the target magnetic property clustering model.

[0143] Among them, the determination of the target posterior probability can be based on expert experience, historical data or other reliable reference information. By comparing the initial posterior probability with the target posterior probability, the accuracy of the current Gaussian mixture model parameters can be evaluated and adjusted accordingly. At the same time, based on the degree of correlation between the attribute feature training samples and the maximum magnetoresistance effect, the update of the model parameters can be made more in line with the needs of practical applications, and the accuracy and reliability of cluster analysis can be improved. When updating the model parameters, optimization algorithms such as gradient descent and Newton's method can be used to ensure the efficiency and accuracy of parameter updates. Ultimately, a Gaussian mixture model that converges and meets the needs of practical applications is obtained, that is, the target magnetic property clustering model. This model can accurately perform cluster analysis on the eigenvalue matrix of the target class resistor material, revealing the magnetic property characteristics and internal laws of the material, and providing strong support for further research and application of the material.

[0144] In an optional embodiment, after clustering the eigenvalues ​​in the eigenvalue matrix to obtain the magnetic property analysis results of the target class resistor material, the magnetic property analysis results can also be physically verified for rationality. The verification includes but is not limited to: checking whether the clustering results are consistent with the known magnetic property laws of the resistor material, evaluating the stability and reliability of the clustering results, and verifying whether the models and methods used in the cluster analysis process are applicable to the target class resistor material. Through physical rationality verification, the accuracy and credibility of the cluster analysis results can be ensured, providing a more reliable basis for subsequent research and application of resistor materials. In specific implementation, a variety of verification methods and means can be adopted, such as comparative analysis, experimental verification, theoretical derivation, etc., to ensure the comprehensiveness and effectiveness of the verification.

[0145] Furthermore, a mapping model between property combinations and the maximum magnetoresistance effect can be established to reveal the specific impact of different property combinations on the maximum magnetoresistance effect, providing a theoretical basis for material optimization design and performance prediction. When constructing a mapping model between property combinations and the maximum magnetoresistance effect, the interactions between multiple properties and their joint contribution to the maximum magnetoresistance effect can be considered. Through machine learning algorithms such as decision trees, random forests, and support vector machines, the complex relationship between property combinations and the maximum magnetoresistance effect can be explored, and a high-precision prediction model can be constructed to quickly evaluate new resistor materials and predict their maximum magnetoresistance effect performance, thereby guiding material screening and optimization.

[0146] The resistor material analysis method provided by the present invention utilizes multi-source magnetic property data of the target resistor material, constructs a property feature matrix and an eigenvalue matrix, and then performs cluster analysis on each eigenvalue in the eigenvalue matrix, thereby being able to comprehensively and quickly obtain the magnetic property analysis results of the target resistor material. This not only improves the analysis efficiency, but also can more accurately capture the complex behavior of the material under different conditions, meeting the application needs of analyzing the maximum magnetoresistance effect of resistor materials.

[0147] Figure 7 FIG. 1 is a schematic structural diagram of an embodiment of a resistance material analysis device according to the present invention. Figure 7 As shown, the device includes:

[0148] The data acquisition module 710 is used to construct a property feature matrix using multi-source magnetic property data of the target type resistor material;

[0149] A feature acquisition module 720 is used to construct an eigenvalue matrix based on the eigenvalues ​​between related attribute features in the attribute feature matrix;

[0150] The material analysis module 730 is used to perform cluster analysis on each eigenvalue in the eigenvalue matrix to obtain the magnetic property analysis results of the target type resistor material.

[0151] In an optional embodiment, the data acquisition module 710 includes:

[0152] The data acquisition submodule is used to obtain the magnetic property test data and magnetic property theoretical data of the target type resistor material. The multi-source magnetic property data includes magnetic property test data and magnetic property theoretical data.

[0153] The data processing submodule is used to process the magnetic property test data and magnetic property theoretical data to obtain target magnetic property data;

[0154] The attribute feature matrix construction submodule is used to construct the attribute feature matrix based on the target magnetic attribute data.

[0155] In an optional embodiment, the data processing submodule includes:

[0156] A data acquisition unit, used to acquire the magnetic property test data after data alignment and missing data and abnormal data in the magnetic property theoretical data;

[0157] A data cleaning unit is used to clean the magnetic property test data and magnetic property theoretical data based on missing data and abnormal data to obtain magnetic property cleaned data;

[0158] The data standardization unit is used to perform data standardization processing on the magnetic property cleaning data to obtain target magnetic property data.

[0159] In an optional embodiment, the feature acquisition module 720 includes:

[0160] An attribute feature acquisition submodule is used to acquire multiple attribute features that are attribute-related and / or attribute-negatively correlated in the attribute feature matrix;

[0161] A covariance matrix construction submodule is used to construct a covariance matrix between multiple attribute features that are attribute-correlated and / or attribute-negatively correlated;

[0162] The eigenvalue matrix construction submodule is used to construct the eigenvalue matrix based on the eigendecomposition values ​​of the covariance matrix.

[0163] In an optional embodiment, the eigenvalue matrix construction submodule includes:

[0164] An eigenvalue decomposition unit is used to perform eigenvalue decomposition on the covariance matrix to obtain eigendecomposition values;

[0165] The eigenvalue matrix construction unit is used to construct an eigenvalue matrix based on eigenvectors corresponding to target eigendecomposition values ​​in the eigendecomposition values.

[0166] In an optional implementation, the eigenvalue matrix construction unit includes:

[0167] The cumulative variance contribution rate calculation subunit is used to calculate the cumulative variance contribution rate of each eigendecomposition value;

[0168] A projection matrix sub-construction unit is used to construct a projection matrix based on eigenvectors corresponding to target eigendecomposition values ​​whose cumulative variance contributions are greater than a contribution threshold;

[0169] The eigenvalue matrix subunit is used to construct an eigenvalue matrix based on the attribute feature matrix and the projection matrix.

[0170] In an optional implementation manner, the eigenvalue matrix construction unit further includes:

[0171] The characteristic curve drawing subunit is used to draw the characteristic curve of the characteristic decomposition value as the number of principal components changes;

[0172] A projection matrix construction subunit is used to construct a projection matrix based on eigenvectors corresponding to target eigendecomposition values ​​whose slopes in the characteristic curve are greater than a slope threshold;

[0173] The eigenvalue matrix construction subunit is used to construct the eigenvalue matrix based on the attribute feature matrix and the projection matrix.

[0174] In an optional embodiment, the material analysis module 730 includes:

[0175] A cluster analysis submodule is used to perform cluster analysis on each eigenvalue in the eigenvalue matrix to obtain a clustering result of the magnetic properties of each resistor material in the target class of resistor materials and / or a clustering result of the magnetic correlation between target class materials;

[0176] The magnetic property analysis submodule is used to obtain the magnetic property analysis results of the target type resistor material based on the magnetic property clustering results and / or the magnetic correlation clustering results.

[0177] In an optional embodiment, the material analysis module 730 is specifically configured to perform cluster analysis on each eigenvalue in the eigenvalue matrix using a target magnetic property clustering model to obtain a magnetic property analysis result of the target type resistor material.

[0178] In an optional embodiment, the material analysis module 730 further includes:

[0179] The training sample acquisition submodule is used to obtain the attribute feature training samples of the target class resistor material;

[0180] The posterior probability calculation submodule is used to input the attribute feature training samples into the Gaussian mixture model to obtain the initial posterior probability that each attribute feature training sample belongs to the corresponding Gaussian component. The model parameters of the Gaussian mixture model include the initial mean vector, the initial covariance matrix and the initial feature weights.

[0181] The target model acquisition submodule is used to update the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain the target magnetic property clustering model.

[0182] In an optional embodiment, the target model acquisition submodule includes:

[0183] An expectation maximization value calculation unit is used to calculate the expectation maximization value of each attribute feature training sample, and the expectation maximization value is used to characterize the contribution degree of each attribute feature training sample to each Gaussian component in the Gaussian mixture model;

[0184] A model parameter adjustment unit is used to adjust the initial mean vector, initial covariance matrix and initial feature weight according to the expected maximization value to obtain updated mean vector, covariance matrix and feature weight;

[0185] The first model acquisition unit is used to iteratively execute the step of calculating the expected maximization value of each attribute feature training sample until the model parameters of the Gaussian mixture model converge, thereby obtaining a target magnetic attribute clustering model.

[0186] In an optional embodiment, the target model acquisition submodule further includes:

[0187] The posterior probability single comparison element is used to compare the initial posterior probability with the target posterior probability to obtain the comparison result;

[0188] The second model acquisition unit is used to update the model parameters of the Gaussian mixture model to the target model parameters based on the difference between the initial posterior probability and the target posterior probability, and the degree of correlation between each attribute feature training sample and the maximum magnetoresistance effect, if the comparison result indicates that the initial posterior probability is less than the target posterior probability, so as to obtain the target magnetic attribute clustering model.

[0189] For the description of the features in the embodiment corresponding to the resistance material analysis device, reference can be made to the relevant description of the embodiment corresponding to the resistance material analysis method, which will not be repeated here.

[0190] The embodiment of the present application also provides a computer device, such as Figure 8 As shown, it includes a memory 810 and a processor 820. The memory 810 stores a computer program, and the processor 820 is configured to run the computer program to execute the steps in any of the above-mentioned resistance material analysis method embodiments.

[0191] The computer device further includes a communication interface 830 for the computer device to communicate with other devices or a communication network.

[0192] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above-mentioned resistance material analysis method embodiments when running.

[0193] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0194] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any of the above-mentioned resistance material analysis method embodiments are implemented.

[0195] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above-mentioned resistance material analysis method embodiments are implemented.

[0196] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0197] The above describes in detail the resistance material analysis method, apparatus, device, and product provided by this application. Specific examples are used herein to illustrate the principles and implementation methods of this application. The above examples are intended only to facilitate understanding of the method and its core concept. It should be noted that those skilled in the art may make various improvements and modifications to this application without departing from the principles of this application, and such improvements and modifications fall within the scope of protection of the claims of this application.

Claims

1. A method for analyzing a resistance material, characterized in that: include: Utilizing multi-source magnetic property data of the target resistive material to construct a property feature matrix, the multi-source magnetic property data includes magnetic property test data and magnetic property theoretical data associated with the maximum magnetoresistance effect, including carrier concentration, mobility, and crystal structure; constructing an eigenvalue matrix based on eigenvalues ​​between related attribute features in the attribute feature matrix; Performing cluster analysis on each eigenvalue in the eigenvalue matrix to obtain magnetic property analysis results of the target resistor material, including electron / hole type energy band classification results; The constructing of an eigenvalue matrix based on the eigenvalues ​​between related attribute features in the attribute feature matrix includes: Acquire multiple attribute features that are attribute-related and / or attribute-negatively correlated in the attribute feature matrix; Constructing a covariance matrix between a plurality of attribute features that are attribute-related and / or attribute-negatively correlated; Performing eigenvalue decomposition on the covariance matrix to obtain eigendecomposition values; Calculate the cumulative variance contribution rate of each eigendecomposition value; Constructing a projection matrix based on eigenvectors corresponding to target eigendecomposition values ​​whose cumulative variance contribution is greater than a contribution threshold; Constructing the eigenvalue matrix based on the attribute feature matrix and the projection matrix; The cluster analysis is performed on each eigenvalue in the eigenvalue matrix to obtain the magnetic property analysis result of the target type resistor material, including: A target magnetic property clustering model is used to perform cluster analysis on each eigenvalue in the eigenvalue matrix to obtain a magnetic property analysis result of the target class resistor material, wherein the magnetic property analysis result includes a magnetic property clustering result of each resistor material and a magnetic correlation clustering result between target class materials, and the magnetic correlation clustering result includes an electron type energy band classification result, a hole type energy band classification result, and a multi-band classification result. The target magnetic property clustering model is obtained based on a Gaussian mixture model and an initial posterior probability. The model parameters of the Gaussian mixture model include an initial mean vector, an initial covariance matrix, and an initial characteristic weight. The initial covariance matrix is ​​used to measure the physical correlation between the energy band structure and the carrier mobility. The obtaining of the target magnetic property clustering model includes: Obtaining a training sample of attribute characteristics of the target type resistor material; Inputting the attribute feature training samples into a Gaussian mixture model to obtain an initial posterior probability that each attribute feature training sample belongs to a corresponding Gaussian component; updating the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain the target magnetic property clustering model; The model of the initial posterior probability that each attribute feature training sample belongs to the corresponding Gaussian component is: The update model of the initial feature weight is: The update model of the initial mean vector is: The update model of the initial covariance matrix is: in, is the probability density function of Gaussian distribution; x j is the attribute feature training sample; i is the i-th Gaussian component, is the initial feature weight; is the initial covariance matrix; is the initial mean vector; Represents the contribution of attribute feature training sample j to the i-th Gaussian component; α i Dynamically adjust the weights for the degree of correlation between the target class resistive material properties and the maximum magnetoresistance effect. It is an indicator of the maximum magnetoresistance effect associated with the attribute feature training sample.

2. The method according to claim 1, characterized in that The method of constructing a property feature matrix by using multi-source magnetic property data of the target resistor material includes: Obtain magnetic property test data and magnetic property theoretical data of target resistor materials; Processing the magnetic property test data and the magnetic property theoretical data to obtain target magnetic property data; The attribute feature matrix is ​​constructed based on the target magnetic attribute data.

3. The method according to claim 2, characterized in that The processing of the magnetic property test data and the magnetic property theoretical data to obtain target magnetic property data includes: Acquiring missing data and abnormal data from the aligned magnetic property test data and the magnetic property theoretical data; performing data cleaning on the magnetic property test data and the magnetic property theoretical data based on the missing data and the abnormal data to obtain magnetic property cleaned data; The magnetic property cleaning data is subjected to data standardization processing to obtain the target magnetic property data.

4. The method according to claim 1, wherein The constructing the eigenvalue matrix based on the eigenvectors corresponding to the target eigendecomposition values ​​in the eigendecomposition values ​​includes: Draw a characteristic curve of the eigendecomposition value as the number of principal components changes; Constructing a projection matrix based on eigenvectors corresponding to target eigendecomposition values ​​whose slopes in the characteristic curve are greater than a slope threshold; The eigenvalue matrix is ​​constructed based on the attribute feature matrix and the projection matrix.

5. The method according to claim 1, wherein The updating of the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain the target magnetic property clustering model includes: Calculating an expectation maximization value of each attribute feature training sample, where the expectation maximization value is used to characterize the contribution of each attribute feature training sample to each Gaussian component in the Gaussian mixture model; Adjusting the initial mean vector, initial covariance matrix, and initial feature weights according to the expected maximization value to obtain updated mean vector, covariance matrix, and feature weights; The step of calculating the expected maximization value of each attribute feature training sample is iteratively performed until the model parameters of the Gaussian mixture model converge, thereby obtaining the target magnetic attribute clustering model.

6. The method according to claim 1, characterized in that The updating of the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain the target magnetic property clustering model includes: Comparing the initial posterior probability with the target posterior probability to obtain a comparison result; If the comparison result indicates that the initial posterior probability is less than the target posterior probability, the model parameters of the Gaussian mixture model are updated to the target model parameters based on the difference between the initial posterior probability and the target posterior probability, and the degree of correlation between each of the attribute feature training samples and the maximum magnetoresistance effect, to obtain the target magnetic attribute clustering model.

7. A resistance material analysis device, characterized in that: include: A data acquisition module is used to construct a property feature matrix using multi-source magnetic property data of the target class resistor material, wherein the multi-source magnetic property data includes magnetic property test data and magnetic property theoretical data associated with the maximum magnetoresistance effect, and the magnetic property test data and magnetic property theoretical data include carrier concentration, mobility, and crystal structure; A feature acquisition module is used to construct an eigenvalue matrix based on the eigenvalues ​​between related attribute features in the attribute feature matrix, including: Acquire multiple attribute features that are attribute-related and / or attribute-negatively correlated in the attribute feature matrix; Constructing a covariance matrix between a plurality of attribute features that are attribute-related and / or attribute-negatively correlated; Performing eigenvalue decomposition on the covariance matrix to obtain eigendecomposition values; Calculate the cumulative variance contribution rate of each eigendecomposition value; Constructing a projection matrix based on eigenvectors corresponding to target eigendecomposition values ​​whose cumulative variance contribution is greater than a contribution threshold; Constructing the eigenvalue matrix based on the attribute feature matrix and the projection matrix; The material analysis module is used to perform cluster analysis on each eigenvalue in the eigenvalue matrix to obtain the magnetic property analysis results of the target type resistor material, including the electron / hole type energy band classification results, including: A target magnetic property clustering model is used to perform cluster analysis on each eigenvalue in the eigenvalue matrix to obtain a magnetic property analysis result of the target class resistor material, wherein the magnetic property analysis result includes a magnetic property clustering result of each resistor material and a magnetic correlation clustering result between target class materials, the magnetic correlation clustering result includes an electronic energy band classification result, a hole energy band classification result, and a multi-band classification result, including an electronic energy band classification result or a hole energy band classification result, the target magnetic property clustering model is obtained based on a Gaussian mixture model and an initial posterior probability, and the model parameters of the Gaussian mixture model include an initial mean vector, an initial covariance matrix, and an initial characteristic weight, and the initial covariance matrix is ​​used to measure the physical correlation between the band structure and the carrier mobility; The obtaining of the target magnetic property clustering model includes: Obtaining a training sample of attribute characteristics of the target type resistor material; Inputting the attribute feature training samples into a Gaussian mixture model to obtain an initial posterior probability that each attribute feature training sample belongs to a corresponding Gaussian component; updating the model parameters of the Gaussian mixture model based on the initial posterior probability to obtain the target magnetic property clustering model; The model of the initial posterior probability that each attribute feature training sample belongs to the corresponding Gaussian component is: The update model of the initial feature weight is: The update model of the initial mean vector is: The update model of the initial covariance matrix is: in, is the probability density function of Gaussian distribution; x j is the attribute feature training sample; i is the i-th Gaussian component, is the initial feature weight; is the initial covariance matrix; is the initial mean vector; Represents the contribution of attribute feature training sample j to the i-th Gaussian component; α i Dynamically adjust the weights for the degree of correlation between the target class resistive material properties and the maximum magnetoresistance effect. It is an indicator of the maximum magnetoresistance effect associated with the attribute feature training sample.

8. A computer device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the resistance material analysis method according to any one of claims 1 to 6 when executing the computer program.

9. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the steps of the resistive material analysis method according to any one of claims 1 to 6.

Images