Unsupervised multi-view feature selection method and device, electronic equipment and storage medium

Abstract

The application provides a kind of unsupervised multi-view feature selection method, device, electronic equipment and storage medium, applied to feature selection technical field.The method comprises: obtaining the multi-view feature of target object;Through the multi-view feature, an unsupervised multi-view feature selection model is constructed, and a feature selection matrix is obtained by solving the unsupervised multi-view feature selection model;According to the feature selection matrix, determine the score of each feature, and carry out feature selection based on the feature score.

Description

Technical Field

[0001] This invention relates to the field of feature selection technology, and in particular to an unsupervised multi-view feature selection method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the development of information technology, it is often possible to acquire large amounts of multi-view data on target objects, such as various feature representations of images and remote sensing data from multiple sensors. Although this multi-view data can provide a more comprehensive description of the target object, the complex relationships between the data and the high dimensionality of the data significantly increase the computational and storage burden. Therefore, it is necessary to filter the multi-view data.

[0003] In existing technologies, feature selection can be performed based on a trained supervised model. However, since the training samples of such supervised models need to contain label information, the process of labeling the information requires a great deal of effort. Summary of the Invention

[0004] This invention provides an unsupervised multi-view feature selection method, apparatus, electronic device, and storage medium to solve the problem that the process of labeling information during the training of feature selection models in the prior art requires a lot of effort.

[0005] This invention provides an unsupervised multi-view feature selection method, comprising: acquiring multi-view features of a target object; constructing an unsupervised multi-view feature selection model using the multi-view features, and solving the unsupervised multi-view feature selection model to obtain a feature selection matrix; determining the score of each feature according to the feature selection matrix, and performing feature selection based on the feature scores.

[0006] According to the unsupervised multi-view feature selection method provided by the present invention, the objective expression of the unsupervised multi-view feature selection model is:

[0007]

[0008] stP=Φ(P (1) ,…,P (m) ),FF T =I;

[0009] in, Let m be the number of views of the multi-view feature, and d be the number of views of the multi-view feature. v Let v be the feature dimension of the viewpoint v, and n be the number of training samples; Let c be the mapping matrix, and c be the total number of categories, ||·|| 2,1 For l 2,1 Norm; P (v) ∈R n×nLet P be the self-representation coefficient of the viewpoint v. The operation of the function Φ is to first convert P... (1) ,…,P (m) Stack the tensors into an n×n×m third-order tensor, then rotate them into an n×m×n third-order tensor P∈R. n×m×n Its inverse function Let λ1, λ2, λ3, and λ4 be the tensor nuclear norm, L be the Laplacian matrix uniform across all perspectives, I be the identity matrix, and λ1, λ2, λ3, and λ4 be the model parameters.

[0010] According to the present invention, an unsupervised multi-view feature selection method is provided, wherein constructing an unsupervised multi-view feature selection model through the multi-view features includes: initializing the target expression and fixing the variables in the target expression to obtain a first objective function, and solving for the feature selection matrix W based on the first objective function. (a) 1≤a≤m; Given that the first objective function satisfies the preset conditions, the variables in the first objective function are fixed to obtain the second objective function, and the model parameters P are solved based on the second objective function. (b) 1≤b≤m; Given that the second objective function satisfies the preset conditions, the variables in the second objective function are fixed to obtain the third objective function, and the model parameters F are solved based on the third objective function; the first objective function, the second objective function, and the third objective function are iteratively solved until the preset conditions are met, and the feature selection matrix W is obtained. (a) The preset condition is that the function is in a convergent state or the number of iterations reaches a preset number.

[0011] According to the unsupervised multi-view feature selection method provided by the present invention, the first objective function is:

[0012]

[0013] The second objective function is:

[0014]

[0015] The third objective function is:

[0016]

[0017] According to the present invention, an unsupervised multi-view feature selection method is provided, wherein determining the score of each feature based on the feature selection matrix and performing feature selection based on the feature score includes: calculating the l2 norm of each row vector of the feature selection matrix and using the l2 norm as the feature score of the corresponding row vector; and selecting features of a preset number of dimensions in descending order of feature scores.

[0018] The present invention also provides an unsupervised multi-view feature selection device, comprising: an acquisition module and a processing module; the acquisition module is used to acquire multi-view features of a target object; the processing module is used to construct an unsupervised multi-view feature selection model through the multi-view features, and solve the unsupervised multi-view feature selection model to obtain a feature selection matrix; determine the score of each feature according to the feature selection matrix, and perform feature selection based on the feature scores.

[0019] According to the present invention, an unsupervised multi-view feature selection device is provided, wherein the objective expression of the unsupervised multi-view feature selection model is:

[0020]

[0021] stP=Φ(P (1) ,…,P (m) ),FF T =I;

[0022] in, Let m be the number of views of the multi-view feature, and d be the number of views of the multi-view feature. v Let v be the feature dimension of the viewpoint v, and n be the number of training samples; Let c be the mapping matrix, and c be the total number of categories, ||·|| 2,1 For l 2,1 Norm; P (v) ∈R n×n Let P be the self-representation coefficient of the viewpoint v. The operation of the function Φ is to first convert P... (1) ,…,P (m) Stack the tensors into an n×n×m third-order tensor, then rotate them into an n×m×n third-order tensor P∈R. n×m×n Its inverse function Let λ1, λ2, λ3, and λ4 be the tensor nuclear norm, L be the Laplacian matrix uniform across all perspectives, I be the identity matrix, and λ1, λ2, λ3, and λ4 be the model parameters.

[0023] According to an unsupervised multi-view feature selection device provided by the present invention, the processing module is used to initialize the target expression, fix the variables in the target expression to obtain a first objective function, and solve for the feature selection matrix W based on the first objective function. (a) 1≤a≤m; Given that the first objective function satisfies the preset conditions, the variables in the first objective function are fixed to obtain the second objective function, and the model parameters P are solved based on the second objective function. (b)1≤b≤m; Given that the second objective function satisfies the preset conditions, the variables in the second objective function are fixed to obtain the third objective function, and the model parameters F are solved based on the third objective function; the first objective function, the second objective function, and the third objective function are iteratively solved until the preset conditions are met, and the feature selection matrix W is obtained. (a) The preset condition is that the function is in a convergent state or the number of iterations reaches a preset number.

[0024] According to the unsupervised multi-view feature selection device provided by the present invention, the first objective function is:

[0025]

[0026] The second objective function is:

[0027]

[0028] The third objective function is:

[0029]

[0030] According to an unsupervised multi-view feature selection device provided by the present invention, the processing module is used to calculate the l2 norm of each row vector of the feature selection matrix and use the l2 norm as the feature score of the corresponding row vector; and select features of a preset number of dimensions in descending order of feature scores.

[0031] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the above-described unsupervised multi-view feature selection methods.

[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described unsupervised multi-view feature selection methods.

[0033] The unsupervised multi-view feature selection method, apparatus, electronic device, and storage medium provided by this invention can acquire multi-view features of a target object; construct an unsupervised multi-view feature selection model using the multi-view features, and solve the unsupervised multi-view feature selection model to obtain a feature selection matrix; determine the score of each feature based on the feature selection matrix, and perform feature selection based on the feature scores. Through this scheme, since an unsupervised multi-view feature selection model can be constructed using multi-view features, and feature selection is performed based on the feature selection matrix of the unsupervised multi-view feature selection model, multi-view feature selection can be completed without the need for data label information. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating the unsupervised multi-view feature selection method provided by the present invention;

[0036] Figure 2 This is a schematic diagram of the structure of the unsupervised multi-view feature selection device provided by the present invention;

[0037] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0039] It should be noted that in the embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or design options. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0040] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of the present invention is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0041] To facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish the same or similar items with essentially the same function and effect. Those skilled in the art can understand that the terms "first" and "second" are not intended to limit the quantity or execution order.

[0042] The embodiments of the present invention have been described for illustrative purposes. It should be understood that the present invention may be implemented in other ways not specifically shown in the accompanying drawings.

[0043] The above implementation method will be described in detail below with reference to specific embodiments and accompanying drawings.

[0044] like Figure 1 As shown, this embodiment of the invention provides an unsupervised multi-view feature selection method, which can be applied to an unsupervised multi-view feature selection device. The unsupervised multi-view feature selection method may include steps S101-S103:

[0045] S101. An unsupervised multi-view feature selection device acquires multi-view features of the target object.

[0046] Optionally, the aforementioned multi-perspective features refer to the features of the target object in different dimensions.

[0047] Optionally, the target object can be an image or a photographed object. The multi-view features of the target object can include Gabor features describing the texture information of the same image, gray-level co-occurrence matrix (GLCM) features, and color features, etc. Alternatively, for multiple images of the same photographed object acquired by multiple sensors and multiple bands, the features of the multiple images can be extracted as the multi-view features of the photographed object.

[0048] Optionally, after obtaining the multi-view features of the target object, the multi-view features can be preprocessed. The preprocessing operation may include at least one of the following: denoising and normalization.

[0049] S102. The unsupervised multi-view feature selection device constructs an unsupervised multi-view feature selection model through the multi-view features and solves the unsupervised multi-view feature selection model to obtain the feature selection matrix.

[0050] Optionally, the objective expression of the unsupervised multi-view feature selection model is:

[0051]

[0052] stP=Φ(P (1) ,…,P (m) ),FF T =I;

[0053] in, Let m be the number of views of the multi-view feature, and d be the number of views of the multi-view feature. v Let v be the feature dimension of the viewpoint v, and n be the number of training samples; Let c be the mapping matrix, and c be the total number of categories, ||·|| 2,1 For l 2,1 Norm; P (v) ∈R n×n Let P be the self-representation coefficient of the viewpoint v. The operation of the function Φ is to first convert P... (1) ,…,P (m) Stack the tensors into an n×n×m third-order tensor, then rotate them into an n×m×n third-order tensor P∈R. n×m×n Its inverse function Let λ1, λ2, λ3, and λ4 be the tensor nuclear norm, L be the Laplacian matrix uniform across all perspectives, I be the identity matrix, and λ1, λ2, λ3, and λ4 be the model parameters.

[0054] Optionally, the unsupervised multi-view feature selection device can first initialize the target expression and fix the variables in the target expression to obtain a first objective function, and then solve for the feature selection matrix W based on the first objective function. (a)1≤a≤m; Given that the first objective function satisfies the preset conditions, the variables in the first objective function are fixed to obtain the second objective function, and the model parameters P are solved based on the second objective function. (b) 1≤b≤m; Given that the second objective function satisfies the preset conditions, the variables in the second objective function are fixed to obtain the third objective function, and the model parameters F are solved based on the third objective function; the first objective function, the second objective function, and the third objective function are iteratively solved until the preset conditions are met, and the feature selection matrix W is obtained. (a) The preset condition is that the function is in a convergent state or the number of iterations reaches a preset number.

[0055] Specifically, the unsupervised multi-view feature selection device can first initialize the target expression, then fix the variables in the target expression to obtain the first target function:

[0056]

[0057] Taking the derivative of the first objective function so that the derivative equals 0, we obtain the following formulas (1) and (2):

[0058] W (v) ={(X (v) -X (v) P (v) (X) (v) -X (v) P (v) ) T +λ1D (v) +λ2X (v) X (v)T} -1 λ2X (v) F T (1);

[0059] Diagonal matrix: D (v) Diagonal elements:

[0060] in, Representation matrix W (v) The i-th row vector, ||·||2 represents the l2 norm.

[0061] Iterate through formulas (1) and (2) until the first objective function is in a convergent state or the number of iterations reaches the preset number.

[0062] Then, by fixing the variables in the first objective function, we obtain the second objective function:

[0063]

[0064] By introducing an auxiliary variable H into the second objective function and adding the constraint H = P, we obtain the following formula (3):

[0065]

[0066] The augmented Lagrangian function of formula (3) is given by formula (4) below:

[0067]

[0068] Where B and μ are Lagrange multiplier parameters.

[0069] For any given viewpoint, the augmented Lagrangian function can be expressed as the following formula (5):

[0070]

[0071] in,

[0072] Taking the derivative of the above formula (5) until the derivative equals 0, we obtain the following formula (6):

[0073]

[0074] Solving for the auxiliary variable H, the second objective function is updated as follows:

[0075]

[0076] Further conversion yields:

[0077]

[0078] The auxiliary variable H in the second objective function can be obtained using the tensor tubal threshold shrinkage operator;

[0079] The update rules for parameters B and μ in the augmented Lagrangian function can be expressed by the following formula (7):

[0080]

[0081] Iterate the auxiliary variable H, the above formulas (6) and (7) until the second objective function is in a convergent state or the number of iterations reaches the preset number.

[0082] Then, by fixing the variables in the second objective function, we obtain the third objective function:

[0083]

[0084] stFF T =I

[0085] W (v)Substituting the expression into the third objective function, we can obtain formula (8):

[0086]

[0087] stFF T =I

[0088] Among them, Q (v) ={(X (v) -X (v) P (v) (X) (v) -X (v) P (v) ) T +λ1D (v) +λ2X (v) X (v)T} -1 λ2X (v) ; By analyzing the matrix The model parameters F are obtained by performing eigenvalue decomposition and taking the transpose of the matrix formed by the eigenvectors corresponding to the c smallest eigenvalues.

[0089] Finally, by iteratively solving the first, second, and third objective functions until convergence is achieved or the preset number of iterations is reached, the feature selection matrix W can be obtained. (a) .

[0090] S103. The unsupervised multi-view feature selection device determines the score of each feature according to the feature selection matrix and performs feature selection based on the feature scores.

[0091] Optionally, the unsupervised multi-view feature selection device can calculate the l2 norm of each row vector of the feature selection matrix and use the l2 norm as the feature score of the corresponding row vector; and select features of a preset number of dimensions in descending order of feature scores.

[0092] Specifically, based on the feature selection matrix W (a) By calculating m feature selection matrices The l2 norm of each row vector can determine the feature score of the corresponding row vector. Based on the feature scores, all row vectors are sorted, and the d features with the largest feature scores are selected for subsequent tasks, where d represents the preset number.

[0093] It should be noted that, on the one hand, this invention proposes an unsupervised multi-view feature selection method based on tensor low-rank minimization, which considers the correlation, complementarity, and consistency of high-order information from different perspectives. It can select a small number of low-dimensional features from multi-view features without requiring data label information. On the other hand, this invention preserves the global structural information of the data through low-rank representation and preserves the local structural information of the data through manifold regularization. In this way, the global and local information of high-dimensional features can be preserved in the feature selection process, thereby better obtaining valuable features.

[0094] In this embodiment of the invention, since an unsupervised multi-view feature selection model can be constructed using multi-view features, and feature selection is performed based on the feature selection matrix of the unsupervised multi-view feature selection model, the selection of multi-view features can be completed without the need for data label information.

[0095] The foregoing mainly describes the solutions provided by the embodiments of the present invention from a methodological perspective. To achieve the above functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0096] The unsupervised multi-view feature selection method provided in this embodiment of the invention can be executed by an unsupervised multi-view feature selection device or a control module for unsupervised multi-view feature selection within that device. This embodiment of the invention uses an unsupervised multi-view feature selection device executing the unsupervised multi-view feature selection method as an example to illustrate the unsupervised multi-view feature selection device provided in this embodiment of the invention.

[0097] It should be noted that, according to the above method examples, the unsupervised multi-view feature selection device can be divided into functional modules in the embodiments of the present invention. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. Optionally, the module division in the embodiments of the present invention is illustrative and only represents one logical functional division; in actual implementation, there may be other division methods.

[0098] like Figure 2As shown, this embodiment of the invention provides an unsupervised multi-view feature selection device 200. The unsupervised multi-view feature selection device 200 includes an acquisition module 201 and a processing module 202. The acquisition module 201 is used to acquire multi-view features of a target object; the processing module 202 is used to construct an unsupervised multi-view feature selection model using the multi-view features, solve the unsupervised multi-view feature selection model to obtain a feature selection matrix; determine the score of each feature based on the feature selection matrix, and perform feature selection based on the feature scores.

[0099] Optionally, the objective expression of the unsupervised multi-view feature selection model is:

[0100]

[0101] stP=Φ(P (1) ,…,P (m) ),FF T =I;

[0102] in, Let m be the number of views of the multi-view feature, and d be the number of views of the multi-view feature. v Let v be the feature dimension of the viewpoint v, and n be the number of training samples; Let c be the mapping matrix, and c be the total number of categories, ||·|| 2,1 For l 2,1 Norm; P (v) ∈R n×n Let P be the self-representation coefficient of the viewpoint v. The operation of the function Φ is to first convert P... (1) ,…,P (m) Stack the tensors into an n×n×m third-order tensor, then rotate them into an n×m×n third-order tensor P∈R. n×m×n Its inverse function Let λ1, λ2, λ3, and λ4 be the tensor nuclear norm, L be the Laplacian matrix uniform across all perspectives, I be the identity matrix, and λ1, λ2, λ3, and λ4 be the model parameters.

[0103] Optionally, the processing module 202 is used to initialize the target expression, fix the variables in the target expression to obtain a first objective function, and solve for the feature selection matrix W based on the first objective function. (a) 1≤a≤m; Given that the first objective function satisfies the preset conditions, the variables in the first objective function are fixed to obtain the second objective function, and the model parameters P are solved based on the second objective function. (b)1≤b≤m; Given that the second objective function satisfies the preset conditions, the variables in the second objective function are fixed to obtain the third objective function, and the model parameters F are solved based on the third objective function; the first objective function, the second objective function, and the third objective function are iteratively solved until the preset conditions are met, and the feature selection matrix W is obtained. (a) The preset condition is that the function is in a convergent state or the number of iterations reaches a preset number.

[0104] Optionally, the first objective function is:

[0105]

[0106] The second objective function is:

[0107]

[0108] The third objective function is:

[0109]

[0110] Optionally, the processing module 202 is used to calculate the l2 norm of each row vector of the feature selection matrix and use the l2 norm as the feature score of the corresponding row vector; and select a preset number of features in descending order of feature scores.

[0111] In this embodiment of the invention, since an unsupervised multi-view feature selection model can be constructed using multi-view features, and feature selection is performed based on the feature selection matrix of the unsupervised multi-view feature selection model, the selection of multi-view features can be completed without the need for data label information.

[0112] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an unsupervised multi-view feature selection method. This method includes: acquiring multi-view features of a target object; constructing an unsupervised multi-view feature selection model using the multi-view features, and solving the unsupervised multi-view feature selection model to obtain a feature selection matrix; determining a score for each feature based on the feature selection matrix, and performing feature selection based on the feature scores.

[0113] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0114] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the unsupervised multi-view feature selection method provided by the above methods, the method comprising: acquiring multi-view features of a target object; constructing an unsupervised multi-view feature selection model through the multi-view features, and solving the unsupervised multi-view feature selection model to obtain a feature selection matrix; determining a score for each feature according to the feature selection matrix, and performing feature selection based on the feature scores.

[0115] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the aforementioned unsupervised multi-view feature selection methods, the method comprising: acquiring multi-view features of a target object; constructing an unsupervised multi-view feature selection model through the multi-view features, and solving the unsupervised multi-view feature selection model to obtain a feature selection matrix; determining a score for each feature based on the feature selection matrix, and performing feature selection based on the feature scores.

[0116] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0117] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An unsupervised multi-view feature selection method, characterized in that, include: Obtain multi-view features of the target object; An unsupervised multi-view feature selection model is constructed using the aforementioned multi-view features, and the feature selection matrix is ​​obtained by solving the unsupervised multi-view feature selection model. Each feature score is determined based on the feature selection matrix, and feature selection is performed based on the feature scores; The objective expression of the unsupervised multi-view feature selection model is: ; in, For the aforementioned multi-view features, The number of viewpoints for the multi-view feature. From the perspective Feature dimensions, The number of training samples; For the mapping matrix, The total number of categories, for Norm; From the perspective The self-representation coefficients of the function The operation is to first put piled up The third-order tensor, then rotated into The third-order tensor Its inverse function , For the tensor nuclear norm, For a unified Laplace matrix from all perspectives, It is the identity matrix. These are model parameters; The construction of an unsupervised multi-view feature selection model using the multi-view features includes: The objective expression is initialized, and the variables in the objective expression are fixed to obtain a first objective function. The feature selection matrix is ​​then solved based on the first objective function. , ; If the first objective function satisfies the preset conditions, the variables in the first objective function are fixed to obtain the second objective function, and the model parameters are solved based on the second objective function. , ; If the second objective function satisfies the preset conditions, the variables in the second objective function are fixed to obtain the third objective function, and the model parameters are solved based on the third objective function. ; The first objective function, the second objective function, and the third objective function are iteratively solved until the preset condition is met, thus obtaining the feature selection matrix. ; The preset condition is that the function is in a convergent state or the number of iterations reaches a preset number.

2. The unsupervised multi-view feature selection method according to claim 1, characterized in that, The first objective function is: ； The second objective function is: ； The third objective function is: 。 3. The unsupervised multi-view feature selection method according to any one of claims 1-2, characterized in that, The step of determining a score for each feature based on the feature selection matrix and performing feature selection based on the feature scores includes: Calculate the value of each row vector of the feature selection matrix. norm, and the said The norm serves as the feature score for the corresponding row vector; Select a preset number of features from each dimension, ranked by feature score from highest to lowest.

4. An unsupervised multi-view feature selection device, characterized in that, include: Acquisition module and processing module; The acquisition module is used to acquire multi-view features of the target object; The processing module is used to construct an unsupervised multi-view feature selection model through the multi-view features, and solve the unsupervised multi-view feature selection model to obtain a feature selection matrix; determine the score of each feature according to the feature selection matrix, and perform feature selection based on the feature scores; The objective expression of the unsupervised multi-view feature selection model is: ; in, For the aforementioned multi-view features, The number of viewpoints for the multi-view feature. From the perspective Feature dimensions, The number of training samples; For the mapping matrix, The total number of categories, for Norm; From the perspective The self-representation coefficients of the function The operation is to first put piled up The third-order tensor, then rotated into The third-order tensor Its inverse function , For the tensor nuclear norm, For a unified Laplace matrix from all perspectives, It is the identity matrix. These are model parameters; The construction of an unsupervised multi-view feature selection model using the multi-view features includes: The objective expression is initialized, and the variables in the objective expression are fixed to obtain a first objective function. The feature selection matrix is ​​then solved based on the first objective function. , ; If the first objective function satisfies the preset conditions, the variables in the first objective function are fixed to obtain the second objective function, and the model parameters are solved based on the second objective function. , ; If the second objective function satisfies the preset conditions, the variables in the second objective function are fixed to obtain the third objective function, and the model parameters are solved based on the third objective function. ; The first objective function, the second objective function, and the third objective function are iteratively solved until the preset condition is met, thus obtaining the feature selection matrix. ; The preset condition is that the function is in a convergent state or the number of iterations reaches a preset number.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the unsupervised multi-view feature selection method as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the unsupervised multi-view feature selection method as described in any one of claims 1 to 3.

Images