CN117912015A - Method and device for selecting multi-mark image data by utilizing local discriminant model and mark correlation - Google Patents
Method and device for selecting multi-mark image data by utilizing local discriminant model and mark correlation Download PDFInfo
- Publication number
- CN117912015A CN117912015A CN202410070932.7A CN202410070932A CN117912015A CN 117912015 A CN117912015 A CN 117912015A CN 202410070932 A CN202410070932 A CN 202410070932A CN 117912015 A CN117912015 A CN 117912015A
- Authority
- CN
- China
- Prior art keywords
- matrix
- image data
- model
- objective function
- local
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 239000011159 matrix material Substances 0.000 claims abstract description 197
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 238000004590 computer program Methods 0.000 claims description 10
- 238000010276 construction Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 description 74
- 238000010586 diagram Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000000835 fiber Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for selecting multi-mark image data by utilizing a local discriminant model and mark correlation, comprising the following steps: constructing adjacent local clusters based on each piece of image data and the adjacent image data, and defining an inter-class discrete matrix, a local discrete matrix and a cluster allocation matrix of the adjacent local clusters to define and obtain a local discrimination model of each piece of image data and a local discrimination model corresponding to the multi-mark image data; the relation between the mark space and the feature space is projected into a feature selection matrix to obtain a loss model, the relation between the feature selection matrix and a cluster allocation matrix is projected into a mark correlation matrix to obtain a correlation model, l 2,1 norms are applied to the feature selection matrix to obtain a feature selection model, an objective function is built, an alternate iterative optimization algorithm is adopted to solve the objective function to obtain a final feature selection matrix, and a feature subset is determined based on the final feature selection matrix so as to improve the classification performance of the multi-mark image data.
Description
Technical Field
The invention relates to the field of image processing, in particular to a method and a device for selecting multi-mark image data by utilizing a local discriminant model and mark correlation.
Background
One area of concern today is multi-label image data classification, as shown in FIG. 3, where an image may be divided into semantic labels such as "beach", "city" and "party". Early researchers used two immature schemes to deal with the multi-mark image data classification problem, one scheme was to convert the multi-mark into a plurality of single marks, and to perform two classifications on the sample on each single mark, which had the disadvantage that the independence of the marks was assumed to ignore the relationship between the marks, and no auxiliary information could be provided for the marks that are associated with each other; another approach is to arrange and combine multiple tags, using "beach", "city", "party", "beach+city", "city+party" etc. as new classes, which is impractical in most cases, the arrangement and combination will greatly increase the number of classes to consider, and the data in such combined classes is typically sparse. In order to cope with the scenes of the multi-label image data, a multi-label learning framework is generated, and the traditional single-label classification method is modified into a multi-label classification method so as to adapt to the classification characteristics of the multi-label image data.
With the advent of the information age, the number of labels for multi-label images has increased, as has the feature dimension. The presence of redundant and uncorrelated features in the high-dimensional features can enlarge the storage space requirements, impeding the classification performance of the multi-label image data. Feature selection is an efficient way to process high-dimensional data by eliminating redundant and irrelevant features to select features of high discriminant. Existing multi-label feature selection methods can be broadly divided into three major categories: filtration, packaging, and embedded processes. The filtered approach does not rely on any learning algorithm in generating the feature subset and therefore does not allow selection of the more desirable features. The wrapper approach employs an evolutionary algorithm to search for feature subsets, but such approach is typically tens of thousands of iterations, and the overhead of computation time is expensive. The embedded method combines the learning model to generate the feature selection matrix, and the feature selection matrix directly determines feature importance ranking, but aiming at multi-mark data, the current embedded method lacks of mining the internal association information of the data.
Disclosure of Invention
The technical problems mentioned above are solved. The embodiment of the application aims to provide a method and a device for selecting multi-mark image data by utilizing a local discriminant model and mark correlation, which are used for solving the technical problems mentioned in the background art section and selecting salient features to represent original multi-mark image data, so that the classification performance of the multi-mark image data is improved.
In a first aspect, the present invention provides a method for selecting multi-label image data using a local discriminant model and label correlation, comprising the steps of:
acquiring multi-mark image data, wherein the multi-mark image data comprises a feature space and a mark space, each piece of image data is contained in the feature space, and the mark space contains a group of marks associated with each piece of image data;
Constructing adjacent local clusters based on each piece of image data and the adjacent image data, defining an inter-class discrete matrix, a local discrete matrix and a cluster distribution matrix of the adjacent local clusters, maximizing the inter-class discrete matrix and minimizing the local discrete matrix through the cluster distribution matrix, defining a local discriminant model of each piece of image data, and calculating a local discriminant model corresponding to the multi-mark image data according to the local discriminant models of all images in the multi-mark image data;
Projecting the relation between the mark space and the feature space into a feature selection matrix to obtain a loss model, projecting the relation between the feature selection matrix and a cluster allocation matrix into a mark correlation matrix to obtain a correlation model, and applying a normal number of l 2,1 on the feature selection matrix to obtain a feature selection model;
and constructing an objective function based on the loss model, the correlation model, the local discriminant model and the feature selection model, solving the objective function by adopting an alternate iterative optimization algorithm to obtain a final feature selection matrix, and determining a feature subset based on the final feature selection matrix.
Preferably, the feature space is a d-dimensional featureThe marking space is marked with c categoriesWherein n represents the number of image data, each image data X i ε X (1.ltoreq.i.ltoreq.n) is associated with a set of markersY i ε Y, where/(v)Indicating that the j-th mark is related to the i-th image data, and if not, indicating/>
Preferably, the adjacent local clusters are represented asComprising the image data x i and adjacent k-1 sheets of image dataDefining a local clique matrix as/>
Preferably, the inter-class discrete matrix is expressed asThe local discrete matrix is denoted/>The definition is as follows:
wherein, Represented as a data matrix after the local clique matrix has been centred,/>And/>The first scaling marking matrix and the second scaling marking matrix are respectively, and I represents a k x k identity matrix;
The cluster allocation matrix is expressed as Wherein sc is the number of clusters, and the second scaling factor matrix is defined as follows:
Wherein S i∈{0,1}n×k is a selection matrix, which is defined as follows:
wherein F i＝{i,i1,...,ik-1 is an adjacent local group And p and q are indexes.
Preferably, the local discrimination model of the i-th image data is defined as:
the local discriminant model corresponding to the multi-label image data is defined as:
wherein, Is a boundary item, and is obtained by simplifying and rewriting the above formula:
wherein,
Preferably, the loss model is defined as:
wherein, Selecting a matrix for the feature,/>Is a bias vector;
The correlation model is defined as:
wherein, Representing a correlation matrix;
the feature selection model is defined as:
Preferably, an objective function is constructed based on a loss model, a correlation model, a local discriminant model and a feature selection model, and the objective function is solved by adopting an alternate iterative optimization algorithm to obtain a final feature selection matrix, and a feature subset is determined based on the final feature selection matrix, which specifically comprises:
Constructing an objective function as shown in the following formula:
s.t.LTL＝I
Wherein, alpha, beta and gamma are the weights of the correlation model, the local discrimination model and the feature selection model respectively; the process of the alternate iterative optimization algorithm is as follows:
S41, setting w= [ W 1,w2,...,wd]T and xw+ nbT-Y＝[z1,z2,...,zn]T, the first equivalent of the objective function is as follows:
Wherein D and Two diagonal matrices, respectively, with diagonal elements D ii＝1/2||wi||2 and D ii＝1/2||wi||2, respectively
S42, fixing other variables to solve the bias vector b: calculating the derivative of the bias vector b to be equal to 0 for the first equivalent of the objective function to obtain the derivative result of the bias vector b:
S43, fixing other variables to solve for P: substituting the derivative of the bias vector b into a first equivalent of the objective function to obtain a first solution of the objective function as shown in the following formula:
wherein, Is a center matrix, and the derivative of the correlation matrix P is equal to 0 in the first solution of the objective function, so as to obtain the derivative result of the correlation matrix P:
s44, fixing other variables to solve for W: substituting the derivative result of the correlation matrix P into a first solution of the objective function to obtain a second solution of the objective function shown in the following formula:
Because of (I-LL T)(I-LLT)＝(I-LLT), the second solution of the objective function is equivalent to the second equivalent of the objective function as shown in the following equation:
solving the derivative of the feature selection matrix W for the second equivalent of the objective function to be equal to 0, to obtain a third solution of the objective function as shown in the following formula:
wherein, B＝A-αLLT,B＝BT；
S45, fixing other variables to solve a clustering distribution matrix L: converting the second equivalent of the objective function to a third equivalent of the objective function based on a and B as shown in the following formula:
Substituting the derivative of the feature selection matrix W in the third solution of the objective function into the third equivalent of the objective function to obtain a fourth solution of the objective function shown in the following formula:
The fourth solution from the SMW identity ：B-1＝(A-αLLT)-1＝A-1+αA-1L(I-αLTA-1L)-1LTA-1, objective function is rewritten as a fourth equivalent of the objective function as shown in the following equation:
Removing extraneous terms in the fourth equivalent of the objective function, reducing to a fifth equivalent of the objective function as shown in the following equation:
For any three matrices U 1,U2 and U 3,Tr(U1U2U3)＝Tr(U2U3U1), then the fifth equivalent of the objective function is rewritten to the sixth equivalent of the objective function as shown in the following equation:
wherein Δ=i- αa -1 and Decomposing the features in the sixth equivalent of the objective function to solve a cluster allocation matrix L;
repeating the iterative steps S41-S45 to enable the objective function to converge, and obtaining a final feature selection matrix W;
the top ranked features are selected as feature subsets by the final feature selection matrix W in descending order of i W i||2 (i=1→d).
In a second aspect, the present invention provides an apparatus for selecting multi-label image data using a local discriminant model and label correlation, comprising:
A data acquisition module configured to acquire multi-label image data, the multi-label image data including a feature space and a label space, each piece of image data being contained in the feature space, the label space containing a group of labels associated with each piece of image data;
the first model construction module is configured to construct adjacent local groups based on each piece of image data and the adjacent image data thereof, define an inter-class discrete matrix and a local discrete matrix of the adjacent local groups and a cluster allocation matrix, maximize the inter-class discrete matrix and minimize the local discrete matrix through the cluster allocation matrix, define and obtain a local discriminant model of each piece of image data, and calculate and obtain a local discriminant model corresponding to the multi-label image data according to the local discriminant models of all images in the multi-label image data;
The second model construction module is configured to project the relation between the mark space and the feature space into the feature selection matrix to obtain a loss model, project the relation between the feature selection matrix and the cluster allocation matrix into the mark correlation matrix to obtain a correlation model, and apply a normal number of l 2,1 on the feature selection matrix to obtain a feature selection model;
And the solving module is configured to construct an objective function based on the loss model, the correlation model, the local discriminant model and the feature selection model, and adopts an alternate iterative optimization algorithm to solve the objective function to obtain a final feature selection matrix, and determine a feature subset based on the final feature selection matrix.
In a third aspect, the present invention provides an electronic device comprising one or more processors; and storage means for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the method as described in any of the implementations of the first aspect.
In a fourth aspect, the present invention provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method as described in any of the implementations of the first aspect.
Compared with the prior art, the invention has the following beneficial effects:
(1) The method for selecting the multi-mark image data by utilizing the local discriminant model and the mark correlation combines the local discriminant model corresponding to each image data, classifies similar image data into the same cluster, and constructs the local discriminant model corresponding to the multi-mark image data.
(2) The method for selecting the multi-label image data by utilizing the local discriminant model and the label correlation utilizes the clustering result to explore the label correlation, constructs a correlation model, and is beneficial to reducing the influence of immature label information.
(3) According to the method for selecting the multi-mark image data by utilizing the local discriminant model and the mark correlation, provided by the invention, the l 2,1 norm regularization term is added into the feature selection matrix to obtain the feature selection model, an objective function is constructed based on the loss model, the correlation model and the local discriminant model corresponding to the multi-mark image data, and the objective function is solved by adopting an alternate iterative optimization algorithm, so that the optimization problem of the objective function can be effectively solved, the final feature selection matrix is accurately solved, and the feature with significance of selection is determined as a feature subset.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is an exemplary device frame pattern to which an embodiment of the present application may be applied;
FIG. 2 is a flow chart of a method for selecting multi-label image data using a local discriminant model and label correlation according to an embodiment of the present application;
FIG. 3 is a schematic diagram of multi-label image data using a local discriminant model and a label correlation selection method for multi-label image data according to an embodiment of the present application;
FIG. 4 is a flow chart of a method for selecting multi-label image data using a local discriminant model and label correlation according to an embodiment of the present application;
FIG. 5 is a schematic diagram of an apparatus for selecting multi-label image data using a local discriminant model and label correlation according to an embodiment of the present application;
Fig. 6 is a schematic structural diagram of a computer device suitable for use in implementing an embodiment of the application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
FIG. 1 illustrates an exemplary device architecture 100 to which embodiments of the present application may be applied to select a multi-labeled image data device using a local discriminant model and label correlation.
As shown in fig. 1, the apparatus architecture 100 may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.
The user may interact with the server 105 via the network 104 using the first terminal device 101, the second terminal device 102, the third terminal device 103, to receive or send messages, etc. Various applications, such as a data processing class application, a file processing class application, and the like, may be installed on the terminal device one 101, the terminal device two 102, and the terminal device three 103.
The first terminal device 101, the second terminal device 102 and the third terminal device 103 may be hardware or software. When the first terminal device 101, the second terminal device 102, and the third terminal device 103 are hardware, they may be various electronic devices, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like. When the first terminal apparatus 101, the second terminal apparatus 102, and the third terminal apparatus 103 are software, they can be installed in the above-listed electronic apparatuses. Which may be implemented as multiple software or software modules (e.g., software or software modules for providing distributed services) or as a single software or software module. The present invention is not particularly limited herein.
The server 105 may be a server that provides various services, such as a background data processing server that processes files or data uploaded by the terminal device one 101, the terminal device two 102, and the terminal device three 103. The background data processing server can process the acquired file or data to generate a processing result.
It should be noted that, the method for selecting the multi-label image data by using the local discriminant model and the label correlation provided in the embodiment of the present application may be performed by the server 105, or may be performed by the first terminal device 101, the second terminal device 102, or the third terminal device 103, and accordingly, the apparatus for selecting the multi-label image data by using the local discriminant model and the label correlation may be provided in the server 105, or may be provided in the first terminal device 101, the second terminal device 102, or the third terminal device 103.
It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation. In the case where the processed data does not need to be acquired from a remote location, the above-described apparatus architecture may not include a network, but only a server or terminal device.
Fig. 2 shows a method for selecting multi-label image data using a local discriminant model and label correlation, according to an embodiment of the present application, comprising the steps of:
S1, acquiring multi-mark image data, wherein the multi-mark image data comprises a feature space and a mark space, each piece of image data is contained in the feature space, and the mark space contains a group of marks associated with each piece of image data.
In a particular embodiment, the feature space is a d-dimensional featureThe tag space is/>, with c class tagsWhere n represents the number of image data, each image data X i ε X (1.ltoreq.i.ltoreq.n) is associated with a set of labels/>Y i ε Y, where/(v)Indicating that the j-th mark is related to the i-th image data, and if not, indicating/>
Specifically, the multi-label image data is as shown in FIG. 3, and first the common symbols of the multi-label image data are given, and the definition of the functions to be used is given below, for any one matrixThe trace of Z is denoted as Tr (Z), the transpose of Z is denoted as Z T, and the Frobenius norm of Z is defined as/>The l 2,1 norm of Z is defined asWhere Z ij represents the ith row and jth column elements of matrix Z.
S2, constructing adjacent local clusters based on each piece of image data and the adjacent image data, defining an inter-class discrete matrix, a local discrete matrix and a cluster allocation matrix of the adjacent local clusters, maximizing the inter-class discrete matrix and minimizing the local discrete matrix through the cluster allocation matrix, defining a local discriminant model of each piece of image data, and calculating a local discriminant model corresponding to the multi-mark image data according to the local discriminant models of all images in the multi-mark image data.
In a particular embodiment, the adjacent local cliques are represented asComprises the image data x i and adjacent k-1 pieces of image data/>Defining a local clique matrix as/>
In a particular embodiment, the inter-class discrete matrix is represented asThe local discrete matrix is denoted/>The definition is as follows:
wherein, Represented as a data matrix after the local clique matrix has been centred,/>And/>The first scaling marking matrix and the second scaling marking matrix are respectively, and I represents a k x k identity matrix;
The cluster allocation matrix is expressed as Wherein sc is the number of clusters, and the second scaling factor matrix is defined as follows:
Wherein S i∈{0,1}n×k is a selection matrix, which is defined as follows:
wherein F i＝{i,i1,...,ik-1 is an adjacent local group And p and q are indexes.
Specifically, each image data and its adjacent image data are constructed into an adjacent local cluster, and a cluster allocation matrix is obtained by using the local discriminant model of all the adjacent local clusters, and the cluster allocation matrix can classify all the image data into a plurality of clusters.
In a specific embodiment, the local discriminant model of the ith image data is defined as:
the local discriminant model corresponding to the multi-label image data is defined as:
wherein, Is a boundary item, and is obtained by simplifying and rewriting the above formula:
wherein,
Specifically, referring to FIG. 4, the matrix is assigned by clustering such that the matrix is discrete between classesMaximization and local discrete matrix/>Thus, a local discriminant model LD i is defined for each image data x i, and the local discriminant models LD i corresponding to all image data are summed together to globally integrate a global local discriminant model LD. In the local discriminant model LD/>Is a boundary term that is used to avoid overfitting of the local discriminant model.
S3, projecting the relation between the mark space and the feature space into a feature selection matrix to obtain a loss model, projecting the relation between the feature selection matrix and a cluster allocation matrix into a mark correlation matrix to obtain a correlation model, and applying a normal number of l 2,1 on the feature selection matrix to obtain a feature selection model.
In a specific embodiment, the loss model is defined as:
wherein, Selecting a matrix for the feature,/>Is a bias vector;
The correlation model is defined as:
wherein, Representing a correlation matrix;
the feature selection model is defined as:
Specifically, the relation between the image data marking space and the characteristic space is projected to a characteristic selection matrix through a loss model, a clustering-based marking correlation matrix is obtained by utilizing the characteristic selection matrix and a clustering distribution matrix, and the relation between the characteristic selection matrix and the clustering distribution matrix is projected to the marking correlation matrix so as to explore marking correlation. To efficiently perform feature selection, the l 2,1 norm is applied to the sparse feature selection matrix. By applying the l 2,1 norm on the feature selection matrix, the performance of the model is more robust and generalizable.
And S4, constructing an objective function based on the loss model, the correlation model, the local discrimination model and the feature selection model, solving the objective function by adopting an alternate iterative optimization algorithm to obtain a final feature selection matrix, and determining a feature subset based on the final feature selection matrix.
In a specific embodiment, step S4 specifically includes:
Constructing an objective function as shown in the following formula:
s.t.LTL＝I
Wherein, alpha, beta and gamma are the weights of the correlation model, the local discrimination model and the feature selection model respectively; the process of the alternate iterative optimization algorithm is as follows:
S41, setting w= [ W 1,w2,...,wd]T and xw+ nbT-Y＝[z1,z2,...,zn]T, the first equivalent of the objective function is as follows:
Wherein D and Two diagonal matrices, respectively, with diagonal elements D ii＝1/2||wi||2 and D ii＝1/2||wi||2, respectively/>
S42, fixing other variables to solve the bias vector b: calculating the derivative of the bias vector b to be equal to 0 for the first equivalent of the objective function to obtain the derivative result of the bias vector b:
S43, fixing other variables to solve for P: substituting the derivative of the bias vector b into a first equivalent of the objective function to obtain a first solution of the objective function as shown in the following formula:
wherein, Is a center matrix, and the derivative of the correlation matrix P is equal to 0 in the first solution of the objective function, so as to obtain the derivative result of the correlation matrix P:
s44, fixing other variables to solve for W: substituting the derivative result of the correlation matrix P into a first solution of the objective function to obtain a second solution of the objective function shown in the following formula:
Because of (I-LL T)(I-LLT)＝(I-LLT), the second solution of the objective function is equivalent to the second equivalent of the objective function as shown in the following equation:
solving the derivative of the feature selection matrix W for the second equivalent of the objective function to be equal to 0, to obtain a third solution of the objective function as shown in the following formula:
/>
wherein, B＝A-αLLT,B＝BT；
S45, fixing other variables to solve a clustering distribution matrix L: converting the second equivalent of the objective function to a third equivalent of the objective function based on a and B as shown in the following formula:
Substituting the derivative of the feature selection matrix W in the third solution of the objective function into the third equivalent of the objective function to obtain a fourth solution of the objective function shown in the following formula:
The fourth solution from the SMW identity ：B-1＝(A-αLLT)-1＝A-1+αA-1L(I-αLTA-1L)-1LTA-1, objective function is rewritten as a fourth equivalent of the objective function as shown in the following equation:
Removing extraneous terms in the fourth equivalent of the objective function, reducing to a fifth equivalent of the objective function as shown in the following equation:
For any three matrices U 1,U2 and U 3,Tr(U1U2U3)＝Tr(U2U3U1), then the fifth equivalent of the objective function is rewritten to the sixth equivalent of the objective function as shown in the following equation:
wherein Δ=i- αa -1 and Decomposing the features in the sixth equivalent of the objective function to solve a cluster allocation matrix L;
repeating the iterative steps S41-S45 to enable the objective function to converge, and obtaining a final feature selection matrix W;
the top ranked features are selected as feature subsets by the final feature selection matrix W in descending order of i W i||2 (i=1→d).
Specifically, based on the steps, an objective function is constructed to guide feature selection, and the objective function is solved through a designed alternate iterative optimization algorithm. The alpha in the objective function explores the influence of the correlation of the high-order marks, the beta controls the contribution of the local discriminant model, and the gamma adjusts the sparsity of the feature selection matrix W. And selecting the features ranked at the front by utilizing the final feature selection matrix obtained by solving to characterize the original image data. In one particular embodiment, the top 20% ranked features are selected as feature subsets in descending order of ||w i||2 (i=1→d). Wherein the feature selection mean W is a matrix of d x c, d is the feature number, c is the label number, corresponding to the importance of each feature to the labels, and by calculating the sum of each row, i.e. the importance of the feature to all labels, all features are calculated sequentially, there is a value of all feature importance, and features can be ranked and feature subsets are selected according to the value of all feature importance, where 20% is the empirical value, and other values can be selected in other embodiments.
The above steps S1-S4 do not necessarily represent the order between steps, but the step symbols indicate that the order between steps is adjustable.
The quality of the feature subset selected by the method for selecting multi-labeled image data using the local discriminant model and the label correlation as proposed by the embodiments of the present application is tested by MLKNN classifier. The average performance of 2 data sets (Image and Scene) on 2 indicators (Hamming loss and Micro-F1) was recorded by five-fold cross-validation, as shown in table 1. The performance of the Method (MFSLL) for selecting multi-label image data by using the local discriminant model and the label correlation according to the embodiment of the application is compared with that of the other 5 comparison methods in the above 2 indexes, wherein the lower the value of Hamming loss, the better the performance, the higher the value of Micro-F1, and the better the performance, and the results are shown in Table 2.
Table 12 detailed characteristics of multiple marker image datasets
Table 2MLKNN Performance of MFSLL on 2 indicators with 5 comparison methods under classifier
With further reference to fig. 5, as an implementation of the method shown in the above figures, the present application provides an embodiment of an apparatus for selecting multi-label image data using a local discriminant model and label correlation, which corresponds to the embodiment of the method shown in fig. 2, and which is particularly applicable to various electronic devices.
The embodiment of the application provides a device for selecting multi-mark image data by utilizing a local discriminant model and mark correlation, which comprises the following steps:
A data acquisition module 1 configured to acquire multi-label image data including a feature space containing each image data and a label space containing a group of labels associated with each image data;
A first model construction module 2 configured to construct an adjacent local group based on each image data and the adjacent image data thereof, define an inter-class discrete matrix and a local discrete matrix of the adjacent local group, and a cluster allocation matrix, maximize the inter-class discrete matrix and minimize the local discrete matrix through the cluster allocation matrix, define a local discriminant model of each image data, and calculate a local discriminant model corresponding to the multi-label image data according to the local discriminant models of all images in the multi-label image data;
The second model construction module 3 is configured to project the relation between the mark space and the feature space into the feature selection matrix to obtain a loss model, project the relation between the feature selection matrix and the cluster allocation matrix into the mark correlation matrix to obtain a correlation model, and apply a normal number of l 2,1 on the feature selection matrix to obtain a feature selection model;
And the solving module 4 is configured to construct an objective function based on the loss model, the correlation model, the local discriminant model and the feature selection model, and adopts an alternate iterative optimization algorithm to solve the objective function to obtain a final feature selection matrix, and determines a feature subset based on the final feature selection matrix.
Referring now to fig. 6, there is illustrated a schematic diagram of a computer apparatus 600 suitable for use in an electronic device (e.g., a server or terminal device as illustrated in fig. 1) for implementing an embodiment of the present application. The electronic device shown in fig. 6 is only an example and should not be construed as limiting the functionality and scope of use of the embodiments of the application.
As shown in fig. 6, the computer apparatus 600 includes a Central Processing Unit (CPU) 601 and a Graphics Processor (GPU) 602, which can perform various appropriate actions and processes according to programs stored in a Read Only Memory (ROM) 603 or programs loaded from a storage section 609 into a Random Access Memory (RAM) 604. In the RAM 604, various programs and data required for the operation of the computer device 600 are also stored. The CPU 601, GPU602, ROM 603, and RAM 604 are connected to each other through a bus 605. An input/output (I/O) interface 606 is also connected to the bus 605.
The following components are connected to the I/O interface 606: an input portion 607 including a keyboard, a mouse, and the like; an output portion 608 including a speaker, such as a Liquid Crystal Display (LCD), etc.; a storage portion 609 including a hard disk and the like; and a communication section 610 including a network interface card such as a LAN card, a modem, or the like. The communication section 610 performs communication processing via a network such as the internet. The drive 611 may also be connected to the I/O interface 606 as needed. A removable medium 612 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 611 as necessary, so that a computer program read out therefrom is mounted into the storage section 609 as necessary.
In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method shown in the flowcharts. In such embodiments, the computer program may be downloaded and installed from a network via the communication portion 610, and/or installed from the removable medium 612. The above-described functions defined in the method of the present application are performed when the computer program is executed by a Central Processing Unit (CPU) 601 and a Graphics Processor (GPU) 602.
It should be noted that the computer readable medium according to the present application may be a computer readable signal medium or a computer readable medium, or any combination of the two. The computer readable medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor apparatus, device, or means, or a combination of any of the foregoing. More specific examples of the computer-readable medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution apparatus, device, or apparatus. In the present application, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may be any computer readable medium that is not a computer readable medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution apparatus, device, or apparatus. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based devices which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules involved in the embodiments of the present application may be implemented in software or in hardware. The described modules may also be provided in a processor.
As another aspect, the present application also provides a computer-readable medium that may be contained in the electronic device described in the above embodiment; or may exist alone without being incorporated into the electronic device. The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to: acquiring multi-mark image data, wherein the multi-mark image data comprises a feature space and a mark space, each piece of image data is contained in the feature space, and the mark space contains a group of marks associated with each piece of image data; constructing adjacent local clusters based on each piece of image data and the adjacent image data, defining an inter-class discrete matrix, a local discrete matrix and a cluster distribution matrix of the adjacent local clusters, maximizing the inter-class discrete matrix and minimizing the local discrete matrix through the cluster distribution matrix, defining a local discriminant model of each piece of image data, and calculating a local discriminant model corresponding to the multi-mark image data according to the local discriminant models of all images in the multi-mark image data; projecting the relation between the mark space and the feature space into a feature selection matrix to obtain a loss model, projecting the relation between the feature selection matrix and a cluster allocation matrix into a mark correlation matrix to obtain a correlation model, and applying a normal number of l 2,1 on the feature selection matrix to obtain a feature selection model; and constructing an objective function based on the loss model, the correlation model, the local discriminant model and the feature selection model, solving the objective function by adopting an alternate iterative optimization algorithm to obtain a final feature selection matrix, and determining a feature subset based on the final feature selection matrix.
The above description is only illustrative of the preferred embodiments of the present application and of the principles of the technology employed. It will be appreciated by persons skilled in the art that the scope of the application referred to in the present application is not limited to the specific combinations of the technical features described above, but also covers other technical features formed by any combination of the technical features described above or their equivalents without departing from the inventive concept described above. Such as the above-mentioned features and the technical features disclosed in the present application (but not limited to) having similar functions are replaced with each other.
Claims (10)
1. A method for selecting multi-label image data using a local discriminant model and label correlation, comprising the steps of:
Acquiring multi-mark image data, wherein the multi-mark image data comprises a feature space and a mark space, each piece of image data is contained in the feature space, and the mark space contains a group of marks associated with each piece of image data;
constructing adjacent local clusters based on each piece of image data and the adjacent image data, defining an inter-class discrete matrix, a local discrete matrix and a cluster allocation matrix of the adjacent local clusters, maximizing the inter-class discrete matrix and minimizing the local discrete matrix through the cluster allocation matrix, defining a local discriminant model of each piece of image data, and calculating the local discriminant model corresponding to the multi-mark image data according to the local discriminant models of all images in the multi-mark image data;
Projecting the relation between the mark space and the feature space into a feature selection matrix to obtain a loss model, projecting the relation between the feature selection matrix and a cluster allocation matrix into a mark correlation matrix to obtain a correlation model, and applying a normal of 2,1 on the feature selection matrix to obtain a feature selection model;
and constructing an objective function based on the loss model, the correlation model, the local discrimination model and the feature selection model, solving the objective function by adopting an alternate iterative optimization algorithm to obtain a final feature selection matrix, and determining a feature subset based on the final feature selection matrix.
2. The method for selecting multi-label image data using local discriminant models and label correlation of claim 1, wherein said feature space is a d-dimensional featureThe marking space is marked with c categoriesWherein n represents the number of image data, each image data X i ε X (1.ltoreq.i.ltoreq.n) is associated with a set of markersY i ε Y, where/(v)Indicating that the j-th mark is related to the i-th image data, and if not, indicating/>
3. The method for selecting multi-label image data using a local discriminant model and label correlation of claim 2, wherein said neighboring local cliques are represented asComprising the image data x i and adjacent k-1 sheets of image dataDefining a local clique matrix as/>
4. A method for selecting multi-label image data using local discriminant models and label correlation as claimed in claim 3, wherein said inter-class discrete matrix is expressed asThe local discrete matrix is expressed as/>The definition is as follows:
wherein, Represented as a data matrix after the local clique matrix is centered,/>And/>The first scaling marking matrix and the second scaling marking matrix are respectively, and I represents a k x k identity matrix;
the cluster allocation matrix is expressed as Wherein sc is a cluster number, and the definition of the second scaling flag matrix is as follows:
Wherein S i∈{0,1}n×k is a selection matrix, which is defined as follows:
Wherein F i＝{i,i1,…,ik-1 is an adjacent local group And p and q are indexes.
5. The method for selecting multi-label image data using local discriminant models and label correlation of claim 4, wherein the local discriminant model for the i-th image data is defined as:
the local discriminant model corresponding to the multi-label image data is defined as:
wherein, Is a boundary item, and is obtained by simplifying and rewriting the above formula:
wherein,
6. The method for selecting multi-label image data using local discriminant models and label correlation of claim 5, wherein said loss model is defined as:
wherein, Selecting a matrix for the feature,/>Is a bias vector;
The correlation model is defined as:
wherein, Representing a correlation matrix;
the feature selection model is defined as:
7. The method for selecting multi-label image data using local discriminant models and label correlation of claim 6, wherein said constructing an objective function based on said loss model, correlation model, local discriminant model and feature selection model, solving said objective function using an alternating iterative optimization algorithm to obtain a final feature selection matrix, determining a feature subset based on said final feature selection matrix, comprising:
Constructing an objective function as shown in the following formula:
s.t.LTL＝I
Wherein alpha, beta and gamma are weights of the correlation model, the local discrimination model and the feature selection model respectively; the process of the alternate iterative optimization algorithm is as follows:
S41, setting w= [ W 1,w2,...,wd]T and xw+ nbT-Y＝[z1,z2,...,zn]T, the first equivalent of the objective function is as follows:
Wherein D and Two diagonal matrices, respectively, with diagonal elements D ii＝1/2||wi||2 and D ii＝1/2||wi||2, respectively
S42, fixing other variables to solve the bias vector b: calculating the derivative of the offset vector b to be equal to 0 according to the first equivalent expression of the objective function, and obtaining the derivative result of the offset vector b:
S43, fixing other variables to solve for P: substituting the derivative of the bias vector b into a first equivalent of the objective function to obtain a first solution of the objective function shown in the following formula:
wherein, Is a center matrix, and the derivative of the correlation matrix P is equal to 0 by the first solution of the objective function, so as to obtain the derivative result of the correlation matrix P:
s44, fixing other variables to solve for W: substituting the derivative result of the correlation matrix P into a first solution formula of the objective function to obtain a second solution formula of the objective function shown in the following formula:
Because of (I-LL T)(I-LLT)＝(I-LLT), the second solution of the objective function is equivalent to the second equivalent of the objective function as shown in the following equation:
And solving the derivative of the feature selection matrix W for the second equivalent expression of the objective function to be equal to 0, and obtaining a third solving expression of the objective function shown in the following expression:
wherein, B＝A-αLLT,B＝BT；
S45, fixing other variables to solve a clustering distribution matrix L: converting the second equivalent of the objective function to a third equivalent of the objective function based on a and B as shown in the following formula:
substituting the derivative result of the feature selection matrix W in the third solution of the objective function into the third equivalent of the objective function to obtain a fourth solution of the objective function shown in the following formula:
The fourth solution to the objective function according to SMW identity ：B-1＝(A-αLLT)-1＝A-1+αA-1L(I-αLTA-1L)-1LTA-1, is rewritten as a fourth equivalent of the objective function as shown in the following equation:
removing extraneous terms in the fourth equivalent of the objective function, and reducing to a fifth equivalent of the objective function as shown in the following formula:
For any three matrices U 1,U2 and U 3,Tr(U1U2U3)＝Tr(U2U3U1), then the fifth equivalent of the objective function is rewritten as the sixth equivalent of the objective function as shown in the following equation:
wherein Δ=i- αa -1 and Decomposing the features in the sixth equivalent of the objective function to solve a cluster allocation matrix L;
repeating the iterative steps S41-S45 to enable the objective function to be converged, and obtaining the final feature selection matrix W;
and selecting the top-ranked features of the final feature selection matrix W as feature subsets in descending order of ||w i||2 (i=1→d).
8. An apparatus for selecting multi-label image data using a local discriminant model and label correlation, comprising:
A data acquisition module configured to acquire multi-label image data, the multi-label image data including a feature space and a label space, each piece of image data being contained in the feature space, the label space containing a group of labels associated with each piece of image data;
The first model construction module is configured to construct adjacent local groups based on each piece of image data and the adjacent image data thereof, define an inter-class discrete matrix and a local discrete matrix of the adjacent local groups and a cluster allocation matrix, maximize the inter-class discrete matrix and minimize the local discrete matrix through the cluster allocation matrix, define and obtain a local discriminant model of each piece of image data, and calculate and obtain a local discriminant model corresponding to the multi-mark image data according to the local discriminant models of all images in the multi-mark image data;
The second model construction module is configured to project the relation between the mark space and the feature space into a feature selection matrix to obtain a loss model, project the relation between the feature selection matrix and a cluster allocation matrix into a mark correlation matrix to obtain a correlation model, and apply a normal of i 2,1 on the feature selection matrix to obtain a feature selection model;
And the solving module is configured to construct an objective function based on the loss model, the correlation model, the local discriminant model and the feature selection model, and adopts an alternate iterative optimization algorithm to solve the objective function to obtain a final feature selection matrix, and determine a feature subset based on the final feature selection matrix.
9. An electronic device, comprising:
one or more processors;
storage means for storing one or more programs,
When executed by the one or more processors, causes the one or more processors to implement the method of any of claims 1-7.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410070932.7A CN117912015A (en) | 2024-01-18 | 2024-01-18 | Method and device for selecting multi-mark image data by utilizing local discriminant model and mark correlation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410070932.7A CN117912015A (en) | 2024-01-18 | 2024-01-18 | Method and device for selecting multi-mark image data by utilizing local discriminant model and mark correlation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117912015A true CN117912015A (en) | 2024-04-19 |
Family
ID=90697351
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410070932.7A Pending CN117912015A (en) | 2024-01-18 | 2024-01-18 | Method and device for selecting multi-mark image data by utilizing local discriminant model and mark correlation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117912015A (en) |
-
2024
- 2024-01-18 CN CN202410070932.7A patent/CN117912015A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Amid et al. | TriMap: Large-scale dimensionality reduction using triplets | |
| CN111755078B (en) | Drug molecule attribute determination method, device and storage medium | |
| CN112395487B (en) | Information recommendation method and device, computer readable storage medium and electronic equipment | |
| CN114936623B (en) | Aspect-level emotion analysis method integrating multi-mode data | |
| CN109885728B (en) | Video abstraction method based on meta-learning | |
| WO2021248492A1 (en) | Semantic representation of text in document | |
| CN113204660B (en) | Multimedia data processing method, tag identification device and electronic equipment | |
| CN113762050B (en) | Image data processing method, device, equipment and medium | |
| John Joseph et al. | A crowding multi-objective genetic algorithm for image parsing | |
| CN113920296B (en) | Text recognition method and system based on comparative learning | |
| CN115905528A (en) | Event multi-label classification method and device with time sequence characteristics and electronic equipment | |
| Hasan et al. | Bangla font recognition using transfer learning method | |
| CN115909336A (en) | Text recognition method and device, computer equipment and computer-readable storage medium | |
| CN115994239A (en) | Prototype comparison learning-based semi-supervised remote sensing image retrieval method and system | |
| Zhang et al. | Saliency detection via sparse reconstruction and joint label inference in multiple features | |
| Liu et al. | Character recognition algorithm based on fusion probability model and deep learning | |
| CN117912015A (en) | Method and device for selecting multi-mark image data by utilizing local discriminant model and mark correlation | |
| CN111259176B (en) | Cross-modal Hash retrieval method based on matrix decomposition and integrated with supervision information | |
| CN116030295A (en) | Article identification method, apparatus, electronic device and storage medium | |
| CN111984800B (en) | Hash cross-modal information retrieval method based on dictionary pair learning | |
| Guo et al. | The rise of nonnegative matrix factorization: algorithms and applications | |
| WO2022123619A1 (en) | Learning system, learning method, and program | |
| Saini et al. | Kannadares-next: A deep residual network for Kannada numeral recognition | |
| Liu et al. | Nonnegative sparse locality preserving hashing | |
| Meddad et al. | A hybrid face identification system using a compressed CNN in a big data environment for embedded devices |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |