WO2022206092A1

WO2022206092A1 - Quantum kernel method-based image classification method and apparatus, server, and system

Info

Publication number: WO2022206092A1
Application number: PCT/CN2022/070264
Authority: WO
Inventors: 杜宇轩; 王新彪; 陶大程; 罗勇
Original assignee: 北京沃东天骏信息技术有限公司
Priority date: 2021-03-29
Filing date: 2022-01-05
Publication date: 2022-10-06
Also published as: CN115222948A

Abstract

Disclosed in embodiments of the present disclosure are a quantum kernel method-based image classification method and apparatus, a server, and a medium. A specific embodiment of the method comprises: obtaining an indefinite kernel matrix obtained by processing an image set on the basis of a quantum kernel method, wherein elements in the indefinite kernel matrix are used for representing the association relationship between images in the image set; generating a corrected eigenvalue diagonal matrix on the basis of correction of a negative value in an eigenvalue diagonal matrix corresponding to the indefinite kernel matrix, wherein an eigenvalue in the corrected eigenvalue diagonal matrix is a non-negative value; generating a corrected positive definite matrix according to an eigenvector matrix corresponding to the indefinite kernel matrix and the corrected eigenvalue diagonal matrix; and generating, according to the corrected positive definite matrix, classification information of the images in the image set by using a pre-trained image classification model. This embodiment reduces the adverse effect of noise of a quantum computer on the accuracy of image classification results.

Description

Image classification method, device, server and system based on quantum kernel method

This patent application claims the priority of the Chinese patent application filed on March 29, 2021 with the application number of 202110334853.9 and the invention titled "image classification method, device, server and system based on quantum kernel method", the full text of the application Incorporated into this application by reference.

technical field

Embodiments of the present disclosure relate to the field of computer technology, and in particular, to an image classification method, apparatus, server, and system based on a quantum kernel method.

Background technique

Kernel method is an important class of algorithms for dealing with nonlinear classification problems in machine learning. Its purpose is to correctly classify data with the same attribute. Its core idea is to transform the original linearly inseparable data into linearly separable data in a high-dimensional space through some nonlinear mathematical transformation.

Due to its own properties (superposition and entanglement), quantum computers are able to generate nonlinear transformations that classical computers cannot simulate. Thus, for a specific dataset, quantum machine learning algorithms can achieve better classification results than optimal classical machine learning algorithms. And the larger the sample size used, the better the classification effect of the algorithm will be.

However, the existing quantum kernel methods do not consider some of the limitations of current real quantum computers. For example, in the era of NISQ (Noisy Intermediate-Scale Quantum, medium-scale noisy quantum devices), quantum computers are noisy, And what we get from a quantum computer is not an exact data, but data that obeys a certain probability distribution. This has a huge impact on some properties of the quantum kernel method, which in turn affects how well it can classify data.

SUMMARY OF THE INVENTION

Embodiments of the present disclosure propose an image classification method, apparatus, server, and system based on a quantum kernel method.

In a first aspect, embodiments of the present disclosure provide an image classification method based on a quantum kernel method, the method comprising: acquiring an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method, wherein the elements in the indeterminate kernel matrix It is used to characterize the relationship between the images in the image set; based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, a revised eigenvalue diagonal matrix is generated, wherein the revised eigenvalues The eigenvalues in the diagonal matrix are non-negative; according to the eigenvector matrix corresponding to the indeterminate kernel matrix and the corrected eigenvalue diagonal matrix, the corrected positive definite matrix is generated; according to the corrected positive definite matrix, the pre-trained image is used The classification model generates classification information for each image in the image collection.

In a second aspect, embodiments of the present disclosure provide an image classification device based on a quantum kernel method, the device comprising: an acquisition unit configured to acquire an indeterminate kernel matrix obtained by processing an image set based on the quantum kernel method, wherein, The elements in the indeterminate kernel matrix are used to characterize the relationship between the images in the image set; the correction unit is configured to generate a modified feature based on the correction of the negative values in the diagonal matrix of eigenvalues corresponding to the indeterminate kernel matrix value diagonal matrix, wherein, the eigenvalues in the corrected eigenvalue diagonal matrix are non-negative values; the generating unit is configured to generate according to the eigenvector matrix corresponding to the indeterminate kernel matrix and the corrected eigenvalue diagonal matrix The modified positive definite matrix; the classification unit is configured to generate classification information of each image in the image set by using a pre-trained image classification model according to the modified positive definite matrix.

In a third aspect, an embodiment of the present disclosure provides an image classification system based on a quantum kernel method, the system includes: a quantum computing terminal, configured to acquire an image set; using a preset quantum kernel function to process the image set to generate an indefinite kernel matrix, wherein the elements in the indefinite kernel matrix are used to characterize the relationship between the images in the image set; the indefinite kernel matrix is sent to the classical computing end; the classical computing end is configured to execute the implementation as in the first aspect The method described by either implementation.

In a fourth aspect, an embodiment of the present disclosure provides a server, the server includes: one or more processors; a storage device on which one or more programs are stored; when one or more programs are processed by one or more The processor executes such that the one or more processors implement the method as described in any one of the implementations of the first aspect.

In a fifth aspect, an embodiment of the present disclosure provides a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, implements the method described in any implementation manner of the first aspect.

The image classification method, device, server and system based on the quantum kernel method provided by the embodiments of the present disclosure, by modifying the negative value in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix for image processing obtained based on the quantum kernel method , so that the corrected positive definite matrix generated according to the corrected eigenvalue diagonal matrix meets the kernel matrix requirements of the classical kernel method, which reduces the adverse effect of the noise of the quantum computer on the accuracy of the image classification result. The combination of classification models improves the accuracy of image classification models.

Description of drawings

Other features, objects and advantages of the present disclosure will become more apparent upon reading the detailed description of non-limiting embodiments taken with reference to the following drawings:

FIG. 1 is an exemplary system architecture diagram to which an embodiment of the present disclosure may be applied;

FIG. 2 is a flowchart of one embodiment of a quantum kernel method-based image classification method according to the present disclosure;

3 is a schematic diagram of an application scenario of an image classification method based on a quantum kernel method according to an embodiment of the present disclosure;

4 is a flowchart of an image classification model obtained by training in an embodiment of the image classification method based on the quantum kernel method according to the present disclosure;

5 is a schematic structural diagram of an embodiment of an image classification apparatus based on a quantum kernel method according to the present disclosure;

FIG. 6 is a sequence diagram of interactions between various devices in one embodiment of the quantum kernel method-based image classification system according to the present disclosure.

7 is a schematic structural diagram of an electronic device suitable for implementing embodiments of the present disclosure.

Detailed ways

The present disclosure will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the related invention, but not to limit the invention. In addition, it should be noted that, for the convenience of description, only the parts related to the related invention are shown in the drawings.

It should be noted that the embodiments of the present disclosure and the features of the embodiments may be combined with each other under the condition of no conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

FIG. 1 illustrates an exemplary architecture 100 to which the quantum kernel method-based image classification method or quantum kernel method-based image classification apparatus of the present disclosure may be applied.

As shown in FIG. 1 , the system architecture 100 may include

terminal devices

101 , 102 , 103 ,

networks

104 , 106 and

servers

105 , 107 . The

networks

104, 106 are used as a medium for providing communication links between the

terminal devices

101, 102, 103 and the server 105, and between the server 105 and the server 107, respectively. The

networks

104, 106 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The

terminal devices

101, 102, and 103 interact with the server 105 through the network 104 to receive or send messages and the like. Various communication client applications can be installed on the

terminal devices

101, 102, and 103, such as search applications, image processing applications, and the like.

The

terminal devices

101, 102, and 103 may be hardware or software. When the

terminal devices

101, 102, and 103 are hardware, they can be various electronic devices having display screens and supporting human-computer interaction, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like. When the

terminal devices

101, 102, and 103 are software, they can be installed in the electronic devices listed above. It can be implemented as a plurality of software or software modules (eg, software or software modules for providing distributed services), or can be implemented as a single software or software module. There is no specific limitation here.

The server 105 may be a quantum computer for providing various services, for example, a background server that provides support for image processing applications on the

terminal devices

101 , 102 , and 103 . The background server can analyze and process the received image, generate a processing result (eg, indeterminate kernel matrix), and send the above-mentioned processing result to the server 107 .

The server 107 may be a classic computer for providing various services, for example, a background server that provides support for image processing applications on the

terminal devices

101 , 102 , and 103 . The background server may analyze and process the received indefinite kernel matrix, and generate a processing result (for example, classification information of each image corresponding to the indefinite kernel matrix).

It should be noted that the above images can also be directly stored locally on the server 105, and the server 105 can directly extract the locally stored images and process them.

It should be noted that the server may be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or can be implemented as a single server. When the server is software, it can be implemented as a plurality of software or software modules (for example, software or software modules for providing distributed services), or can be implemented as a single software or software module. There is no specific limitation here.

It should be noted that the image classification method based on the quantum kernel method provided by the embodiments of the present disclosure is generally executed by the server 107 , and accordingly, the image classification apparatus based on the quantum kernel method is generally set in the server 107 .

It should be understood that the numbers of terminal devices, networks and servers in FIG. 1 are merely illustrative. There can be any number of terminal devices, networks and servers according to implementation needs.

Continuing to refer to FIG. 2 , a flow 200 of one embodiment of a quantum kernel method-based image classification method according to the present disclosure is shown. The image classification method based on the quantum kernel method includes the following steps:

Step 201: Obtain an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method.

In this embodiment, the execution body of the image classification method based on the quantum kernel method (the server 107 shown in FIG. 1 ) can obtain the indeterminate kernel obtained by processing the image set based on the quantum kernel method through wired connection or wireless connection matrix. Wherein, the elements in the above-mentioned indeterminate kernel matrix can be used to represent the relationship between the images in the image set. As an example, the above-mentioned relationship may be the inner product between the data obtained after the image is mapped by the quantum kernel function.

In this embodiment, the above-mentioned execution subject may acquire an indeterminate kernel matrix that is pre-stored locally and obtained by processing an image set based on a quantum kernel method. The above-mentioned executive body may also acquire an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method from an electronic device (eg, the quantum computer 105 shown in FIG. 1 ) connected to it in communication.

It should be noted that since the quantum computer itself has noise, it will cause disturbance to the elements in the generated quantum nuclear matrix. Therefore, the actually generated kernel matrix is often an indefinite matrix, that is, a matrix whose eigenvalues contain both positive and negative values.

In some optional implementations of this embodiment, the noise of the quantum computer on which the above quantum kernel method relies is smaller than a first preset threshold.

It should be noted that the inventors found that the noise of the quantum computer used to execute the above quantum kernel method has a great influence on the image classification accuracy when it reaches a certain level. Based on the above-mentioned optional implementations, this solution can control the noise of the quantum computer within the first preset threshold, so that the image classification using the quantum kernel method has a more significant improvement in accuracy compared to the image classification of the classical computer. Effect.

Step 202 , based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, generate a corrected eigenvalue diagonal matrix.

In this embodiment, based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix obtained in the above step 201, the above-mentioned executive body may generate the corrected eigenvalue diagonal matrix in various ways. Wherein, the eigenvalues in the above-mentioned corrected eigenvalue diagonal matrix are non-negative values.

In this embodiment, the above-mentioned execution body may perform eigendecomposition (A=QΣQ ⁻¹ ) on the indefinite kernel matrix (for example, matrix A) obtained in the above-mentioned step 201 to generate a diagonal matrix (for example, diagonal matrix) including at least one eigenvalue matrix Σ) and eigenvector matrix (such as matrix Q). The eigenvalues (eg, λ _i =Σ _ii ) in the above-mentioned diagonal matrix include both positive and negative values. For the generated negative eigenvalues in the above-mentioned diagonal matrix, the above-mentioned executive body may modify the above-mentioned negative eigenvalues into non-negative eigenvalues in various ways. As an example, the above-mentioned execution body may correct the above-mentioned negative eigenvalues to corresponding non-negative eigenvalues according to a preset correction value correspondence table. The above correction value correspondence table may be used to represent the correspondence between negative values and corrected non-negative eigenvalues.

In some optional implementations of this embodiment, the above-mentioned execution body may correct the negative values in the diagonal matrix of eigenvalues corresponding to the indeterminate kernel matrix obtained in the above step 201 to preset values, and generate the corrected eigenvalues diagonal matrix.

In these implementations, the above-mentioned preset value may be a non-negative value. As an example, the above executive body may replace the negative values in the eigenvalue diagonal matrix corresponding to the above indefinite kernel matrix with a preset value (for example, 0.01), and keep other eigenvalues unchanged, thereby generating a modified eigenvalue diagonal matrix .

In some optional implementations of this embodiment, the above-mentioned execution body may correct the negative value in the diagonal matrix of eigenvalues corresponding to the indeterminate kernel matrix obtained in the above step 201 to the absolute value of the above-mentioned negative value, and generate a modified The eigenvalue diagonal matrix of .

In these implementations, as an example, the eigenvalue diagonal matrix corresponding to the above indefinite kernel matrix may include "-0.3" and "-0.8". The above-mentioned execution body can correct the above-mentioned "-0.3" to "0.3", and the above-mentioned "-0.8" to "0.8", and keep other eigenvalues unchanged, thereby generating the corrected eigenvalue diagonal matrix.

In some optional implementations of this embodiment, the above-mentioned execution body may generate a corrected diagonal matrix of eigenvalues according to the following steps:

In the first step, the eigenvalue with the smallest numerical value in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix is determined as the target value.

In these implementations, as an example, the eigenvalue diagonal matrix corresponding to the above indefinite kernel matrix may be diag(-0.3, 0.6, 0.1, -0.5). The above executive body may determine "-0.5" as the target value.

In the second step, for the eigenvalue in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, the sum of the absolute value of the eigenvalue and the target value is determined as the corrected eigenvalue.

In these implementations, as an example, for the eigenvalues "-0.3, 0.6, 0.1, -0.5" in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, the above-mentioned executive body can compare the absolute value of each eigenvalue and the target value The sum of "0.5" is determined as the modified eigenvalue corresponding to the original eigenvalue, that is, "0.2, 1.1, 0.6, 0" is determined as the modified eigenvalue for "-0.3, 0.6, 0.1, -0.5".

In the third step, a modified diagonal matrix of eigenvalues is generated according to the determined modified eigenvalues.

In these implementation manners, the above-mentioned execution body may form the modified eigenvalues into a modified eigenvalue diagonal matrix according to the arrangement of eigenvalues in the original eigenvalue diagonal matrix. As an example, the above-mentioned execution body may generate the corrected eigenvalue diagonal matrix diag(0.2, 1.1, 0.6, 0).

Step 203: Generate a modified positive definite matrix according to the eigenvector matrix corresponding to the indefinite kernel matrix and the modified eigenvalue diagonal matrix.

In this embodiment, according to the eigenvector matrix corresponding to the indeterminate kernel matrix obtained in step 201 and the modified diagonal matrix of eigenvalues generated in step 202, the above-mentioned executive body can generate the modified positive definite matrix in various ways. The eigenvector matrix corresponding to the above indefinite kernel matrix is usually an eigenvector matrix obtained by eigendecomposition of the indefinite kernel matrix obtained in the above step 201 . The above-mentioned execution body may multiply the corrected eigenvalue diagonal matrix generated in step 202 by the above-mentioned corresponding eigenvector matrix in a manner matching the above-mentioned eigendecomposition to obtain a corrected matrix. Usually, the modified matrix obtained above is a positive definite matrix.

Step 204 , according to the corrected positive definite matrix, use a pre-trained image classification model to generate classification information of each image in the image set.

In this embodiment, according to the corrected positive definite matrix, the above-mentioned executing subject may generate classification information of each image in the image set by using a pre-trained image classification model. The above-mentioned image classification model is used to represent the correspondence between the positive definite matrix and the classification information of the image. The above image classification model may include various models trained by machine learning methods, such as SVM (Support Vector Machine, support vector machine). The above classification information may include various forms, for example, "0" and "1" are used to represent two different categories.

Continue to refer to FIG. 3 , which is a schematic diagram of an application scenario of the image classification method based on the quantum kernel method according to an embodiment of the present disclosure. In the application scenario of FIG. 3 , the background server 306 can obtain the indeterminate kernel matrix 305 . The backend server 306 may modify the negative values in the diagonal matrix based on the eigenvalues obtained by eigendecomposition of the indefinite kernel matrix 305 to obtain a modified diagonal matrix of eigenvalues. Then, the backend server 306 may generate a modified positive definite matrix 307 according to the eigenvector matrix obtained by eigendecomposition of the indefinite kernel matrix 305 and the modified eigenvalue diagonal matrix. After that, the background server 306 inputs the above-mentioned corrected positive definite matrix 307 to the pre-trained image classification model 308 to generate classification information 309 of each image.

Optionally, the above-mentioned indeterminate kernel matrix 305 can be obtained through the following process: the user can send the image to be processed to the quantum computer 304 through the terminal devices 301 and 302 (such as the image _xi located in the image space x as shown in 303 in FIG. x _j ). The quantum computer 304 can use the quantum kernel function to convert the image of the image space x to a higher dimension to generate an indeterminate kernel matrix 305 . After that, the quantum computer can send the generated indeterminate kernel matrix 305 to the above-mentioned background server 306 .

Currently, one of the existing techniques usually does not take into account some of the limitations of current real quantum computers, resulting in the noise of the quantum computer adversely affecting the classification accuracy of quantum kernel-based methods. However, in the method provided by the above embodiments of the present disclosure, by modifying the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix for image processing obtained based on the quantum kernel method, the modified eigenvalue diagonal matrix The generated corrected positive definite matrix complies with the kernel matrix requirements of the classical kernel method, which reduces the adverse effect of the noise of the quantum computer on the accuracy of the image classification result, and then improves the accuracy of the image classification model through the combination of the quantum kernel method and the classical classification model. .

Further referring to FIG. 4 , it shows a process 400 of training an image classification model in an embodiment of the image classification method based on the quantum kernel method. The process 400 of obtaining an image classification model by training includes the following steps:

Step 401, acquiring training samples.

In this embodiment, the execution subject for training the image classification model may acquire training samples from a local or communicatively connected electronic device. The above-mentioned training samples may include a sample indeterminate kernel matrix obtained by processing the sample image set based on the quantum kernel method and label information corresponding to each sample image in the above-mentioned sample image set. Wherein, the elements in the sample indeterminate kernel matrix may be used to represent the association relationship between the sample images in the sample image set.

In this embodiment, as an example, the total number of sample images included in the sample image set is n, and the dimension of the sample indeterminate kernel matrix may be n×n. The elements in the first row and the first column in the sample indeterminate kernel matrix can be used to represent the inner product of the first sample image after being processed by the quantum kernel function and itself.

In some optional implementations of this embodiment, the above-mentioned sample image set generally belongs to a core-set (core-set) of the original sample image set. The number of sample images included in the above-mentioned sample image set is usually less than the second preset threshold.

It should be noted that the inventor found that the number of sample images included in the image set has a great influence on the image classification accuracy to a certain extent. Different from classical machine learning, the larger the number of samples, the better the learning effect, that is, the higher the accuracy of the classification results. When the dimension of the indeterminate kernel matrix obtained by the quantum kernel method reaches a certain level, the accuracy of the classification results will be reduced compared with the advantages of classical machine learning.

Based on the above-mentioned optional implementation manners, this solution can extract core subsets from the original sample image set through an active learning query strategy, so that the number of sample images contained in the sample image set is less than the second preset threshold, thereby ensuring the use of quantum The accuracy advantage of the kernel method for image classification compared to the classical computer image classification.

In step 402, the sample indeterminate kernel matrix of the training sample is used as input, and the annotation information corresponding to each sample image corresponding to the input sample indeterminate kernel matrix is used as the expected output, and an image classification model is obtained by training.

In this embodiment, the above-mentioned execution body may use the sample indeterminate kernel matrix of the training sample obtained in step 401 as an input, and use the annotation information corresponding to each sample image corresponding to the input sample indeterminate kernel matrix as the expected output, through the machine learning method. Train to get an image classification model. The above-mentioned image classification models may include various classification models based on kernel methods, such as SVM.

It should be noted that the executive body used for training the image classification model and the executive body used for executing the image classification method based on the quantum kernel method may be the same or different, which is not limited herein.

It can be seen from FIG. 4 that the process of obtaining an image classification model by training in the image classification method based on the quantum kernel method in this embodiment embodies the sample indeterminate kernel matrix and the The step of performing model training on the training samples of annotation information corresponding to each sample image in the sample image set. Therefore, the solution described in this embodiment provides a training method of an image classification model based on the quantum kernel method, thereby improving the accuracy of image classification.

Further referring to FIG. 5 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an image classification apparatus based on a quantum kernel method, which is similar to the method embodiment shown in FIG. 2 or FIG. 4 . Correspondingly, the apparatus can be specifically applied to various electronic devices.

As shown in FIG. 5 , the image classification apparatus 500 based on the quantum kernel method provided in this embodiment includes an acquisition unit 501 , a correction unit 502 , a generation unit 503 and a classification unit 504 . Wherein, the acquiring unit 501 is configured to acquire an indeterminate kernel matrix obtained by processing the image set based on the quantum kernel method, wherein the elements in the indeterminate kernel matrix are used to represent the correlation between the images in the image set; the modifying unit 502 , is configured to generate a corrected eigenvalue diagonal matrix based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, wherein the eigenvalues in the corrected eigenvalue diagonal matrix are non-negative value; the generating unit 503 is configured to generate a modified positive definite matrix according to the corresponding eigenvector matrix of the indeterminate kernel matrix and the modified eigenvalue diagonal matrix; the classification unit 504 is configured to, according to the modified positive definite matrix, Use a pre-trained image classification model to generate classification information for each image in the image set.

In this embodiment, in the image classification apparatus 500 based on the quantum kernel method: the specific processing of the acquisition unit 501 , the correction unit 502 , the generation unit 503 and the classification unit 504 and the technical effects brought by them can be implemented with reference to FIG. 2 respectively. The related descriptions of step 201 , step 202 , step 203 and step 204 in the example will not be repeated here.

In some optional implementations of this embodiment, the above-mentioned correcting unit 502 may be further configured to: correct the negative values in the diagonal matrix of eigenvalues corresponding to the indeterminate kernel matrix to preset values, and generate the corrected eigenvalues A diagonal matrix where the default values are non-negative.

In some optional implementations of this embodiment, the above-mentioned modifying unit 502 may be further configured to: modify the negative value in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix to the absolute value of the negative value, and generate a modified Eigenvalue diagonal matrix.

In some optional implementations of this embodiment, the above modification unit 502 may be further configured to: determine the eigenvalue with the smallest numerical value in the diagonal matrix of eigenvalues corresponding to the indeterminate kernel matrix as the target value; for the indeterminate kernel matrix corresponding to The eigenvalues in the diagonal matrix of eigenvalues of .

In some optional implementations of this embodiment, the noise of the quantum computer on which the quantum kernel method is based may be smaller than the first preset threshold.

In some optional implementations of this embodiment, the above-mentioned pre-trained image classification model may be obtained by training through the following steps: acquiring training samples, wherein the training samples include indeterminate samples obtained by processing a sample image set based on a quantum kernel method The kernel matrix and the annotation information corresponding to each sample image in the sample image set; the sample indeterminate kernel matrix of the training sample is used as input, and the annotation information corresponding to each sample image corresponding to the input sample indeterminate kernel matrix is used as the expected output. Get an image classification model.

In some optional implementations of this embodiment, the above-mentioned sample image set may belong to a core subset of the original sample image set. The number of sample images included in the above-mentioned sample image set may be less than the second preset threshold.

In the apparatus provided by the above embodiments of the present disclosure, the correction unit 502 corrects the negative values in the diagonal matrix of eigenvalues corresponding to the indeterminate kernel matrix for image processing obtained by the acquisition unit 501 based on the quantum kernel method, so that the generation of The corrected positive definite matrix generated by the unit 503 according to the corrected eigenvalue diagonal matrix meets the requirements of the kernel matrix of the classical kernel method, which reduces the adverse effect of the noise of the quantum computer on the accuracy of the image classification result. The combination of classification models improves the accuracy of image classification models.

With further reference to FIG. 6, a sequence 600 of interactions between various devices in one embodiment of the quantum kernel method-based image classification method is shown. The image classification system based on the quantum kernel method may include: a quantum computing terminal (eg, the server 105 shown in FIG. 1 ) and a classical computing terminal (eg, the server 107 shown in FIG. 1 ). The above quantum computing terminal can be configured to acquire an image set; use a preset quantum kernel function to process the image set to generate an indeterminate kernel matrix; and send the indeterminate kernel matrix to the classical computing terminal. Wherein, the elements in the above-mentioned indeterminate kernel matrix can be used to represent the relationship between the images in the image set. The above-mentioned classical computing terminal can be configured to implement the image classification method based on the quantum kernel method as described in the foregoing embodiments.

As shown in FIG. 6, in step 601, the quantum computing terminal acquires an image set.

In this embodiment, the quantum computing terminal may acquire the image set from a local or communicatively connected electronic device (eg,

terminal devices

101, 102, 103 shown in FIG. 1 ) through a wired or wireless connection. Wherein, the above-mentioned image set may generally include at least one image to be classified.

In step 602, the quantum computing terminal uses a preset quantum kernel function to process the image set to generate an indeterminate kernel matrix.

In this embodiment, the quantum computing end may use a preset quantum kernel function to process the image set acquired in step 601 to generate an indeterminate kernel matrix. Wherein, the elements in the indeterminate kernel matrix can be used to represent the association relationship between the images in the above-mentioned image set. The above-mentioned association relationship may be consistent with the description in the foregoing embodiments, and details are not repeated here.

In step 603, the quantum computing terminal sends the indeterminate kernel matrix to the classical computing terminal.

In this embodiment, the quantum computing terminal may send the indeterminate kernel matrix generated in step 602 to the classical computing terminal. Among them, the above-mentioned classical computing terminal is the currently widely used conventional computer corresponding to the quantum computer.

In step 604, the classical computing terminal obtains the indeterminate kernel matrix obtained by processing the image set based on the quantum kernel method.

In step 605, based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, the classical computing terminal generates a corrected eigenvalue diagonal matrix.

In step 606, according to the eigenvector matrix corresponding to the indefinite kernel matrix and the corrected eigenvalue diagonal matrix, the classical computing terminal generates a corrected positive definite matrix.

In step 607, according to the corrected positive definite matrix, the classical computing terminal generates classification information of each image in the image set by using the pre-trained image classification model.

The above steps 604 to 607 are respectively consistent with the steps 201 to 204 and their optional implementations in the foregoing embodiment, and the above descriptions for the steps 201 to 204 and their optional implementations are also applicable to the steps 604 to 204. Step 607 is not repeated here.

The image classification system based on the quantum kernel method provided by the above embodiments of the present disclosure performs mapping processing based on the quantum kernel method on the image set through the quantum computing end to generate an indeterminate kernel matrix, and the classical computing end matches the image obtained by the quantum kernel method. The correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix of image processing makes the corrected positive definite matrix generated according to the corrected eigenvalue diagonal matrix conforms to the kernel matrix requirements of the classical kernel method and reduces the quantum computer. Therefore, the accuracy of the image classification model is improved by the combination of the quantum kernel method and the classical classification model.

Referring next to FIG. 7 , it shows a schematic structural diagram of an electronic device (eg, server 107 in FIG. 1 ) 700 suitable for implementing embodiments of the present disclosure. The server shown in FIG. 7 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. 7 , an electronic device 700 may include a processing device (eg, a central processing unit, a graphics processor, etc.) 701 that may be loaded into random access according to a program stored in a read only memory (ROM) 702 or from a storage device 708 Various appropriate actions and processes are executed by the programs in the memory (RAM) 703 . In the RAM 703, various programs and data required for the operation of the electronic device 700 are also stored. The processing device 701, the ROM 702, and the RAM 703 are connected to each other through a bus 704. An input/output (I/O) interface 705 is also connected to bus 704 .

In general, the following devices may be connected to the I/O interface 705: input devices 706 including, for example, a touch screen, touch pad, keyboard, mouse, etc.; output devices including, for example, a Liquid Crystal Display (LCD), speakers, vibrators, etc. 707; storage devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication means 709 may allow electronic device 700 to communicate wirelessly or by wire with other devices to exchange data. While Figure 7 illustrates electronic device 700 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. Each block shown in FIG. 7 can represent one device, and can also represent multiple devices as required.

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 709, or from the storage device 708, or from the ROM 702. When the computer program is executed by the processing device 701, the above-described functions defined in the methods of the embodiments of the present disclosure are executed.

It should be noted that the computer-readable medium described in the embodiments of 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 embodiments of the present 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. Rather, in embodiments of the present disclosure, a computer-readable signal medium may include a data signal in baseband or propagated as part of a carrier wave, carrying computer-readable program code therein. 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 . The program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to: electric wire, optical cable, RF (Radio Frequency, radio frequency), etc., or any suitable combination of the above.

The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or may exist alone without being assembled into the server. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the server, the server: obtains an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method, wherein the indeterminate kernel The elements in the matrix are used to characterize the relationship between the images in the image set; based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, a revised eigenvalue diagonal matrix is generated, where the correction The eigenvalues in the eigenvalue diagonal matrix are non-negative; according to the eigenvector matrix corresponding to the indefinite kernel matrix and the revised eigenvalue diagonal matrix, a revised positive definite matrix is generated; according to the revised positive definite matrix, use The pre-trained image classification model generates classification information for each image in the image collection.

Computer program code for carrying out operations of embodiments of the present disclosure may be written in one or more programming languages, including object-oriented programming languages—such as Java, Smalltalk, C++, or a combination thereof, Also included are conventional procedural programming languages - such as "C", the Python 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 (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 software or hardware. The described unit may also be set in the processor, for example, it may be described as: a processor, including an acquisition unit, a correction unit, a generation unit, and a classification unit. Among them, the names of these units do not constitute a limitation of the unit itself in some cases. For example, the acquisition unit can also be described as "a unit that acquires an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method, where , the elements in the indeterminate kernel matrix are used to characterize the association between the images in the image collection".

The above description is merely a preferred embodiment of the present disclosure and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solution formed by the specific combination of the above-mentioned technical features, and should also cover, without departing from the above-mentioned inventive concept, the above-mentioned Other technical solutions formed by any combination of technical features or their equivalent features. For example, a technical solution is formed by replacing the above-mentioned features with the technical features disclosed in the embodiments of the present disclosure (but not limited to) with similar functions.

Claims

An image classification method based on the quantum kernel method, including:

obtaining an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method, wherein the elements in the indeterminate kernel matrix are used to characterize the correlation between the images in the image set;

Based on the correction of the negative values in the eigenvalue diagonal matrix corresponding to the indefinite kernel matrix, a revised eigenvalue diagonal matrix is generated, wherein the eigenvalues in the revised eigenvalue diagonal matrix are non-negative value;

According to the eigenvector matrix corresponding to the indefinite kernel matrix and the modified eigenvalue diagonal matrix, a modified positive definite matrix is generated;

According to the corrected positive definite matrix, a pre-trained image classification model is used to generate classification information of each image in the image set.
The method according to claim 1, wherein, generating a corrected eigenvalue diagonal matrix based on the correction of negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, comprising:

The negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix are corrected to a preset value, and a corrected eigenvalue diagonal matrix is generated, wherein the preset value is a non-negative value.
The method according to claim 1, wherein, generating a corrected eigenvalue diagonal matrix based on the correction of negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, comprising:

Correcting the negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix to the absolute value of the negative values, and generating the corrected eigenvalue diagonal matrix.
The method according to claim 1, wherein, generating a corrected eigenvalue diagonal matrix based on the correction of negative values in the eigenvalue diagonal matrix corresponding to the indeterminate kernel matrix, comprising:

Determine the eigenvalue with the smallest numerical value in the eigenvalue diagonal matrix corresponding to the indefinite kernel matrix as the target value;

For the eigenvalue in the eigenvalue diagonal matrix corresponding to the indefinite kernel matrix, the sum of the absolute value of the eigenvalue and the target value is determined as the corrected eigenvalue;

Based on the determined corrected eigenvalues, a corrected eigenvalue diagonal matrix is generated.
The method according to claim 1, wherein the noise of the quantum computer on which the quantum kernel method relies is smaller than a first preset threshold.
The method according to any one of claims 1-5, wherein the pre-trained image classification model is obtained by training through the following steps:

Obtaining a training sample, wherein the training sample includes a sample indeterminate kernel matrix obtained by processing a sample image set based on a quantum kernel method and annotation information corresponding to each sample image in the sample image set;

The image classification model is obtained by training the sample indefinite kernel matrix of the training sample as input, and the label information corresponding to each sample image corresponding to the input sample indeterminate kernel matrix as expected output.
The method according to claim 6, wherein the sample image set belongs to a core subset of the original sample image set, and the number of sample images included in the sample image set is less than a second preset threshold.
An image classification device based on quantum kernel method, comprising:

an acquiring unit, configured to acquire an indeterminate kernel matrix obtained by processing an image set based on a quantum kernel method, wherein the elements in the indeterminate kernel matrix are used to characterize the relationship between the images in the image set;

a correction unit, configured to generate a corrected eigenvalue diagonal matrix based on the correction of negative values in the eigenvalue diagonal matrix corresponding to the indefinite kernel matrix, wherein, in the corrected eigenvalue diagonal matrix The eigenvalues of are non-negative;

a generating unit, configured to generate a modified positive definite matrix according to the eigenvector matrix corresponding to the indefinite kernel matrix and the modified eigenvalue diagonal matrix;

The classification unit is configured to use a pre-trained image classification model to generate classification information of each image in the image set according to the modified positive definite matrix.
An image classification system based on the quantum kernel method, including:

The quantum computing end is configured to acquire an image set; use a preset quantum kernel function to process the image set to generate an indeterminate kernel matrix, wherein the elements in the indefinite kernel matrix are used to represent the images in the image set. correlation between images; send the indeterminate kernel matrix to the classical computing terminal;

The classical computing terminal is configured to execute the method according to any one of claims 1-7.
A server that includes:

one or more processors;

a storage device on which one or more programs are stored;

The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-7.
A computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method of any one of claims 1-7.