CN114742132A - Deep multi-view clustering method, system and device based on shared difference learning - Google Patents

Abstract

The embodiment of the disclosure provides a deep multi-view clustering method, system and device based on common difference learning, belonging to the technical field of data processing, and specifically comprising the following steps: establishing a common difference depth multi-view feature learning network; respectively connecting each view of the multi-view data with a common information extraction network and a difference information extraction network; inputting the common information extraction network of all views of the multi-view data into a common information learning module for training until convergence; inputting a common information extraction network and a difference information extraction network of all views of the multi-view data into a difference information learning module, and obtaining the complementary characteristics of each view of the multi-view data through orthogonal constraint; connecting the consistency features and all the complementary features in series to form a multi-view fusion feature; and inputting the multi-view fusion characteristics into a clustering model based on KL divergence for clustering. By the scheme, the clustering effect and the adaptability of the multi-view data under the condition of initial severe imbalance are improved.

Description

Deep multi-view clustering method, system and device based on shared difference learning

technical field

The embodiments of the present disclosure relate to the technical field of data processing, and in particular, to a deep multi-view clustering method, system and device based on shared difference learning.

Background technique

At present, the basic idea of clustering is to divide the samples into several clusters according to the similarity between the samples in the data set, and the similarity between the samples of the same kind is smaller than the similarity between the samples of the different classes. Traditional clustering algorithms are mainly for single-view data, and the data has only one set of features. When the data has multiple sets of features, it is called multi-view data. Multi-view data not only contains more rich and useful information, but also brings redundant information between different views. At present, most multi-view clustering mainly focuses on maximizing the common information of multi-views, ignoring the difference information of each view, that is, the complementary information of multi-view data is not fully mined; the initial multi-view data is seriously unbalanced. In this case, using the existing methods may produce the "cask effect", that is, the common information of all views will move towards the view with the worst initial features, and the features of high-quality views are not fully utilized, which also loses data. Meaning from a multi-view description.

It can be seen that a deep multi-view clustering method based on shared difference learning with high clustering effect and adaptability is urgently needed.

SUMMARY OF THE INVENTION

In view of this, the embodiments of the present disclosure provide a deep multi-view clustering method, system and device based on shared difference learning, which at least partially solves the problems of clustering effect and poor adaptability for utilizing high-quality view features in the prior art. question.

In a first aspect, an embodiment of the present disclosure provides a deep multi-view clustering method based on shared difference learning, including:

Step 1, establishing a common difference depth multi-view feature learning network, wherein the common difference depth multi-view feature learning network includes a depth feature extraction module, a common information learning module and a difference information learning module, and the deep feature extraction module includes common information. Extraction network and differential information extraction network;

Step 2, acquiring multi-view data, and connecting each view of the multi-view data to the common information extraction network and the difference information extraction network respectively;

Step 3, the common information extraction network of all views of the multi-view data is input into the common information learning module for training until convergence, and the consistency feature of the multi-view data is obtained;

Step 4, input the common information extraction network and the difference information extraction network of all views of the multi-view data into the difference information learning module, and obtain the complementary feature of each view of the multi-view data through orthogonal constraints;

Step 5, connecting the consistent feature and all the complementary features in series to form a multi-view fusion feature;

Step 6: Input the multi-view fusion feature into a KL divergence-based clustering model for clustering.

According to a specific implementation of the embodiment of the present disclosure, the shared information learning module includes a generative adversarial network.

According to a specific implementation manner of the embodiment of the present disclosure, the step 3 specifically includes:

Step 3.1, the shared information learning module uses the shared information extraction network on each view as a generator G, and finally obtains M generators;

Step 3.2, pass the feature data generated by the M generators into the discriminator D of the M classification;

In step 3.3, steps 3.1 and 3.2 are repeated until the discriminator cannot distinguish the views corresponding to the feature data, and the consistent feature is obtained.

其中，h_i表示第m个视图中的第i个样本的多视图融合特征，

和

分别表示在视图m上提取到的共有信息和差异信息。According to a specific implementation manner of the embodiment of the present disclosure, the series connection manner of the step 5 is:

where h _i represents the multi-view fusion feature of the i-th sample in the m-th view,

and

represent the common information and difference information extracted on the view m, respectively.

According to a specific implementation manner of the embodiment of the present disclosure, the step 6 specifically includes:

The multi-view fusion feature is input into a KL divergence-based clustering model to iteratively train the common difference depth multi-view feature learning network and the clustering network, and the multi-view data is clustered.

In a second aspect, an embodiment of the present disclosure provides a deep multi-view clustering system based on shared difference learning, including:

The establishment module is used to establish a common difference depth multi-view feature learning network, wherein the common difference depth multi-view feature learning network includes a depth feature extraction module, a common information learning module and a difference information learning module, and the depth feature extraction module includes Common information extraction network and differential information extraction network;

an acquisition module, configured to acquire multi-view data, and connect each view of the multi-view data to the common information extraction network and the difference information extraction network respectively;

The first learning module is configured to input the common information extraction network of all views of the multi-view data into the common information learning module for training until convergence, and obtain the consistency features of the multi-view data;

The second learning module is used to input the common information extraction network and the difference information extraction network of all views of the multi-view data into the difference information learning module, and obtain the complementary features of each view of the multi-view data through orthogonal constraints ;

a fusion module, configured to concatenate the consistent feature and all the complementary features to form a multi-view fusion feature;

The clustering module is used for inputting the multi-view fusion feature into a KL divergence-based clustering model for clustering.

In a third aspect, an embodiment of the present disclosure further provides an electronic device, the electronic device comprising:

at least one processor; and,

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the aforementioned first aspect or any implementation of the first aspect based on Deep multi-view clustering methods for shared differential learning.

The deep multi-view clustering scheme based on common difference learning in the embodiment of the present disclosure includes: step 1, establishing a common difference depth multi-view feature learning network, wherein the common difference depth multi-view feature learning network includes a deep feature extraction module , a common information learning module and a difference information learning module, the deep feature extraction module includes a common information extraction network and a difference information extraction network; Step 2, obtain multi-view data, and connect each view of the multi-view data to all the The common information extraction network and the difference information extraction network; Step 3, the common information extraction network of all views of the multi-view data is input into the common information learning module for training until convergence, and the consistency feature of the multi-view data is obtained. Step 4, the common information extraction network and the difference information extraction network of all views of the multi-view data are input to the difference information learning module, and the complementary feature of each view of the multi-view data is obtained by orthogonal constraints; Step 5 , the consistent feature and all the complementary features are connected in series to form a multi-view fusion feature; step 6, the multi-view fusion feature is input into a clustering model based on KL divergence for clustering.

The beneficial effects of the embodiments of the present disclosure are: through the solution of the present disclosure, the multi-view feature learning and multi-view clustering strategies make full use of the consistency and complementary information of multi-view data, and at the same time reduce redundant information between different views, and improve the Clustering effect and adaptability.

Description of drawings

In order to explain the technical solutions of the embodiments of the present disclosure more clearly, the following briefly introduces the accompanying drawings that need to be used in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

1 is a schematic flowchart of a deep multi-view clustering method based on shared difference learning according to an embodiment of the present disclosure;

2 is a schematic flowchart of another deep multi-view clustering method based on shared difference learning provided by an embodiment of the present disclosure;

3 is a schematic structural diagram of a common difference depth multi-view feature learning network according to an embodiment of the present disclosure;

FIG. 4 is a schematic structural diagram of a deep feature extraction module according to an embodiment of the present disclosure;

5 is a schematic structural diagram of a shared information learning module according to an embodiment of the present disclosure;

6 is a schematic structural diagram of a difference information learning module according to an embodiment of the present disclosure;

7 is a schematic structural diagram of a deep multi-view clustering system based on shared difference learning according to an embodiment of the present disclosure;

FIG. 8 is a schematic diagram of an electronic device according to an embodiment of the present disclosure.

Detailed ways

The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

The embodiments of the present disclosure are described below through specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only some, but not all, embodiments of the present disclosure. The present disclosure can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments may be combined with each other under the condition of no conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure.

It is noted that various aspects of embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms and that any specific structure and/or function described herein is illustrative only. Based on this disclosure, those skilled in the art should appreciate that an aspect described herein may be implemented independently of any other aspects and that two or more of these aspects may be combined in various ways. For example, an apparatus may be implemented and/or a method may be practiced using any number of the aspects set forth herein. Additionally, such an apparatus may be implemented and/or such a method may be practiced using other structure and/or functionality in addition to one or more of the aspects set forth herein.

It should also be noted that the drawings provided in the following embodiments are only illustrative of the basic concept of the present disclosure, and the drawings only show the components related to the present disclosure rather than the number, shape and the number of components in actual implementation. For dimension drawing, the type, quantity and proportion of each component can be changed at will in actual implementation, and the component layout may also be more complicated.

Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, one skilled in the art will understand that the described aspects may be practiced without these specific details.

Embodiments of the present disclosure provide a deep multi-view clustering method based on shared difference learning, and the method can be applied to a multi-view data clustering analysis process in a data processing scenario.

Referring to FIG. 1 , it is a schematic flowchart of a deep multi-view clustering method based on shared difference learning according to an embodiment of the present disclosure. As shown in Figure 1 and Figure 2, the method mainly includes the following steps:

Step 1, establishing a common difference depth multi-view feature learning network, wherein the common difference depth multi-view feature learning network includes a depth feature extraction module, a common information learning module and a difference information learning module, and the deep feature extraction module includes common information. Extraction network and differential information extraction network;

In specific implementation, the common difference depth multi-view feature learning network can be constructed first, so that the common difference depth multi-view feature learning network can make full use of the consistency and complementary information of multi-view data, and at the same time reduce the difference between different views. redundant information, the structure of the common difference depth multi-view feature learning network is shown in Figure 3, at the same time, the common difference depth multi-view feature learning network includes the deep feature extraction module as shown in Figure 4, the common difference depth multi-view feature learning network The information learning module is shown in Figure 5, and the difference information learning module is shown in Figure 6, wherein the deep feature extraction module includes a common information extraction network and a difference information extraction network.

Step 2, acquiring multi-view data, and connecting each view of the multi-view data to the common information extraction network and the difference information extraction network respectively;

d^m表示在第m个视图中样本的特征维度，N表示样本的总数。然后可以将所述多视图数据的每个视图分别连接所述共有信息提取网络和所述差异信息提取网络，则所述共有信息提取网络输出的是每个视图的共有信息，所述差异信息提取网络输出的是每个视图的差异信息。For example, let the multi-view data be X={X ⁽¹⁾ ,X ⁽²⁾ ,...,X ^(M) }, where M represents the total number of views of the data,

d ^m represents the feature dimension of the samples in the mth view, and N represents the total number of samples. Then, each view of the multi-view data can be connected to the common information extraction network and the difference information extraction network respectively, then the common information extraction network outputs the common information of each view, and the difference information extraction network outputs the common information of each view. The output of the network is the difference information for each view.

Specifically, the deep feature extraction model includes: a common information extraction network and a difference information extraction network, which are used to extract the common information of all views and the difference information contained in each view.

It is assumed that the common information extraction sub-network and the difference information extraction sub-network in each view contain n+1 fully connected layers, and the k∈[0,n]th layer contains and p _sk units, respectively. Then for the sample x in the mth view, the output of the kth layer of the shared information network can be expressed as:

和

分别表示共有信息提取子网络中第k层的权值矩阵和偏移向量。

是非线性激活函数，常用的有sigmoid和tanh等。in,

and

respectively represent the weight matrix and offset vector of the kth layer in the common information extraction sub-network.

is a nonlinear activation function, commonly used are sigmoid and tanh.

Meanwhile, for the sample x in the mth view, the output of the kth layer of the disparity information network can be expressed as:

和

分别表示差异信息提取子网络中第k层的权值矩阵和偏移向量。in,

and

respectively represent the weight matrix and offset vector of the kth layer in the difference information extraction sub-network.

可以分别得到其对应的共有信息和差异信息，记作：

(共有信息)和

(差异信息)。So for the ith sample in the mth view

The corresponding common information and difference information can be obtained respectively, denoted as:

(shared information) and

(difference information).

Step 3, the common information extraction network of all views of the multi-view data is input into the common information learning module for training until convergence, and the consistency feature of the multi-view data is obtained;

Optionally, the shared information learning module includes a generative adversarial network.

Further, the step 3 specifically includes:

Step 3.1, the shared information learning module uses the shared information extraction network on each view as a generator G, and finally obtains M generators;

Step 3.2, pass the feature data generated by the M generators into the discriminator D of the M classification;

In step 3.3, steps 3.1 and 3.2 are repeated until the discriminator cannot distinguish the views corresponding to the feature data, and the consistent feature is obtained.

During specific implementation, the shared information learning module uses the deep shared information extraction sub-network on each view as a generator G, and finally obtains M generators. The feature data generated by the M generators are passed into the discriminator D of the M classification. The goal of G is to generate data with a similar distribution, so that the discriminator D cannot distinguish which view the data comes from, while the goal of D is to try to distinguish which view the incoming feature data comes from or through which generator G is generated. Through adversarial learning, the shared information extracted from each view is finally similar enough to form the consistent feature. That is, to maximize the common information in different views. The module objective function is as follows:

表示样本

通过生成器G^m生成的样本，

表示样本

的真实视图标签。

的输出为生成的样本来源于视图m的概率值，即：where G ^m represents the generator (common information extraction network) on the mth view,

represent samples

The samples generated by the generator G ^m ,

represent samples

true view label.

The output of is the probability value that the generated sample originates from view m, namely:

Step 4, input the common information extraction network and the difference information extraction network of all views of the multi-view data into the difference information learning module, and obtain the complementary feature of each view of the multi-view data through orthogonal constraints;

During specific implementation, the common information extraction network and the difference information extraction network of all views of the multi-view data can be input into the difference information learning module, and the correlation between the common information and the difference information is minimized through orthogonal constraints, thereby obtaining Complementary features for each view.

Step 5, connecting the consistent feature and all the complementary features in series to form a multi-view fusion feature;

其中，h_i表示第m个视图中的第i个样本的多视图融合特征，

和

分别表示在视图m上提取到的共有信息和差异信息。On the basis of the above-mentioned embodiment, the series connection method of the step 5 is:

where h _i represents the multi-view fusion feature of the i-th sample in the m-th view,

and

represent the common information and difference information extracted on the view m, respectively.

用

和

分别表示在视图m上提取到的共有信息和差异信息。然后，将所有视图上提取到的共有信息向量和差异信息向量，通过以下方式融合，得到样本i的共有差异信息h_i作为所述多视图融合特征。In specific implementation, for the i-th sample in the m-th view

use

and

represent the common information and difference information extracted on the view m, respectively. Then, the common information vector and the difference information vector extracted from all the views are fused in the following manner to obtain the common difference information h _i of the sample i as the multi-view fusion feature.

Among them, h _c,i represents the common information of all views, which is calculated by the following formula,

Step 6: Input the multi-view fusion feature into a KL divergence-based clustering model for clustering.

Optionally, the step 6 specifically includes:

The multi-view fusion feature is input into a KL divergence-based clustering model to iteratively train the common difference depth multi-view feature learning network and the clustering network, and the multi-view data is clustered.

In specific implementation, the common difference features of multiple views can be input into the KL-based clustering model, the common difference information learning network and the clustering network can be iteratively trained, and the multi-view data can be clustered. The objective function of the deep multi-view clustering algorithm based on shared difference learning is:

L=L _c +λ ₁ L _s +λ ₂ L _clu (8)

Among them, λ ₁ , λ ₂ are balance factors, which are used to adjust the proportion of each part of the loss in the total objective function, and L _clu is the clustering loss, which is calculated by the following formula:

where K is the number of categories of clusters, q _ij is the soft assignment probability that sample i belongs to cluster j, and p _ij is the target probability that sample i belongs to cluster j.

q _ij and p _ij are calculated by:

Among them, u _j is the cluster center of the jth class, and α is a degree of freedom variable. In order to simplify the calculation, its value is fixed to 1.

where f _j is the sum of the soft assignment probabilities of all samples belonging to the jth cluster:

The deep multi-view clustering method based on shared difference learning provided by this embodiment includes two networks through a deep feature extraction sub-module, one of which is used to extract common information, and the other is used to extract difference information. The common information learning module, through the fusion of GAN technology, makes the common information extracted from each view as similar as possible; the difference information learning module, through orthogonal constraints, minimizes the correlation between the common information and the difference information.

Then, the common difference depth multi-view feature learning network is applied to multi-view clustering, and the common difference information extracted by the multi-view feature learning network is passed to the subsequent clustering network, and the common difference information learning network and clustering are trained iteratively. The network achieves the purpose of clustering multi-view data, and improves the clustering effect and adaptability.

Corresponding to the above method embodiments, referring to FIG. 7 , an embodiment of the present disclosure further provides a deep multi-view clustering system 70 based on shared difference learning, including:

The establishment module 701 is used to establish a common difference depth multi-view feature learning network, wherein the common difference depth multi-view feature learning network includes a depth feature extraction module, a common information learning module and a difference information learning module, and the depth feature extraction module Including common information extraction network and differential information extraction network;

an acquisition module 702, configured to acquire multi-view data, and connect each view of the multi-view data to the common information extraction network and the difference information extraction network respectively;

The first learning module 703 is configured to input the common information extraction network of all views of the multi-view data into the common information learning module for training until convergence, and obtain the consistency features of the multi-view data;

The second learning module 704 is configured to input the common information extraction network and the difference information extraction network of all views of the multi-view data into the difference information learning module, and obtain the complementarity of each view of the multi-view data through orthogonal constraints feature;

A fusion module 705, configured to concatenate the consistent feature and all the complementary features to form a multi-view fusion feature;

The clustering module 706 is configured to input the multi-view fusion feature into a KL divergence-based clustering model for clustering.

The system shown in FIG. 7 can correspondingly execute the content in the foregoing method embodiment. For the parts not described in detail in this embodiment, reference is made to the content recorded in the foregoing method embodiment, and details are not repeated here.

Referring to FIG. 8 , an embodiment of the present disclosure further provides an electronic device 80 , where the electronic device includes: at least one processor and a memory communicatively connected to the at least one processor. Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the deep multi-view based on mutual difference learning in the foregoing method embodiments clustering method.

Embodiments of the present disclosure also provide a non-transitory computer-readable storage medium, where the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to perform the learning based on shared differences in the foregoing method embodiments A deep multi-view clustering method.

Embodiments of the present disclosure also provide a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, when the program instructions are executed by a computer, make The computer executes the deep multi-view clustering method based on shared difference learning in the foregoing method embodiments.

Referring next to FIG. 8 , it shows a schematic structural diagram of an electronic device 80 suitable for implementing an embodiment of the present disclosure. The electronic devices in the embodiments of the present disclosure may include, but are not limited to, such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablets), PMPs (portable multimedia players), vehicle-mounted terminals (eg, mobile terminals such as in-vehicle navigation terminals), etc., and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device shown in FIG. 8 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present disclosure.

As shown in FIG. 8 , electronic device 80 may include processing means (eg, central processing unit, graphics processor, etc.) 801 that may be loaded into random access according to a program stored in read only memory (ROM) 802 or from storage means 808 Various appropriate actions and processes are executed by the programs in the memory (RAM) 803 . In the RAM 803, various programs and data necessary for the operation of the electronic device 80 are also stored. The processing device 801 , the ROM 802 , and the RAM 803 are connected to each other through a bus 804 . An input/output (I/O) interface 805 is also connected to bus 804 .

Typically, the following devices can be connected to the I/O interface 805: input devices 806 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; including, for example, a liquid crystal display (LCD), speakers, An output device 807 of a vibrator or the like; a storage device 808 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 809 . Communication means 809 may allow electronic device 80 to communicate wirelessly or by wire with other devices to exchange data. While the figures show electronic device 80 having various means, it should be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided.

In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the method illustrated in the flowchart. In such an embodiment, the computer program may be downloaded and installed from the network via the communication device 809 , or from the storage device 808 , or from the ROM 802 . When the computer program is executed by the processing device 801, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are executed.

It should be noted that the computer-readable medium mentioned above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples of computer readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable Programmable read only memory (EPROM or flash memory), fiber optics, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. In this disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with computer-readable program code embodied thereon. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device . Program code embodied on a computer readable medium may be transmitted using any suitable medium including, but not limited to, electrical wire, optical fiber cable, RF (radio frequency), etc., or any suitable combination of the foregoing.

The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or may exist alone without being assembled into the electronic device.

The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device can execute the relevant steps of the above-mentioned method embodiments.

Alternatively, the above computer-readable medium carries one or more programs, and when the above one or more programs are executed by the electronic device, the electronic device can execute the relevant steps of the above method embodiments.

Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including object-oriented programming languages—such as python, Smalltalk, C++, but also conventional Procedural programming language - such as the "C" language or similar programming language. 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 (eg, using an Internet service provider through Internet connection).

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more logical functions for implementing the specified functions executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the blocks 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 is also noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented in dedicated hardware-based systems that perform the specified functions or operations , or can be implemented in a combination of dedicated hardware and computer instructions.

The units involved in the embodiments of the present disclosure may be implemented in a software manner, and may also be implemented in a hardware manner.

It should be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof.

The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited to this. Any person skilled in the art who is familiar with the technical scope of the present disclosure can easily think of changes or substitutions. All should be included within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims (7)

1. A deep multi-view clustering method based on common difference learning is characterized by comprising the following steps:

step 1, establishing a common difference depth multi-view feature learning network, wherein the common difference depth multi-view feature learning network comprises a depth feature extraction module, a common information learning module and a difference information learning module, and the depth feature extraction module comprises a common information extraction network and a difference information extraction network;

step 2, obtaining multi-view data, and respectively connecting each view of the multi-view data with the common information extraction network and the difference information extraction network;

step 3, inputting the common information of all views of the multi-view data into a common information learning module for training until convergence, and obtaining the consistency characteristic of the multi-view data;

step 4, inputting the common information extraction network and the difference information extraction network of all the views of the multi-view data into a difference information learning module, and obtaining the complementary characteristics of each view of the multi-view data through orthogonal constraint;

step 5, connecting the consistency features and all the complementary features in series to form a multi-view fusion feature;

and 6, inputting the multi-view fusion characteristics into a clustering model based on KL divergence for clustering.

2. The method of claim 1, wherein the common information learning module comprises generating a countermeasure network.

3. The method according to claim 2, wherein the step 3 specifically comprises:

step 3.1, the common information learning module takes a common information extraction network on each view as a generator G to finally obtain M generators;

step 3.2, transmitting the feature data generated by the M generators into a discriminator D of M categories;

and 3.3, repeating the step 3.1 and the step 3.2 until the discriminator cannot distinguish the view corresponding to the characteristic data to obtain the consistency characteristic.

4. The method according to claim 1, wherein the series connection of step 5 is

Wherein h is_iA multi-view fusion feature representing the ith sample in the mth view,

and

respectively representing the common information and the disparity information extracted on view m.

5. The method according to claim 1, wherein the step 6 specifically comprises:

inputting the multi-view fusion characteristics into a KL divergence-based clustering model to iteratively train the common difference depth multi-view characteristic learning network and the clustering network, and clustering the multi-view data.

6. A deep multi-view clustering system based on common difference learning, comprising:

the system comprises an establishing module, a judging module and a judging module, wherein the establishing module is used for establishing a common difference depth multi-view feature learning network, the common difference depth multi-view feature learning network comprises a depth feature extracting module, a common information learning module and a difference information learning module, and the depth feature extracting module comprises a common information extracting network and a difference information extracting network;

the acquisition module is used for acquiring multi-view data and respectively connecting each view of the multi-view data with the common information extraction network and the difference information extraction network;

the first learning module is used for inputting the common information extraction network of all the views of the multi-view data into the common information learning module for training until convergence to obtain the consistency characteristic of the multi-view data;

the second learning module is used for inputting the common information extraction network and the difference information extraction network of all the views of the multi-view data into the difference information learning module, and obtaining the complementary characteristics of each view of the multi-view data through orthogonal constraint;

a fusion module for concatenating the consistent features and all of the complementary features to form a multi-view fusion feature;

and the clustering module is used for inputting the multi-view fusion characteristics into a clustering model based on KL divergence for clustering.

7. An electronic device, characterized in that the electronic device comprises:

at least one processor; and the number of the first and second groups,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the common difference learning based deep multiview clustering method of any of the preceding claims 1-5.

Images