Adaptive frequency contrast learning graph collaborative filtering recommendation method

Abstract

The invention relates to the technical field of machine learning, in particular to an adaptive frequency contrast learning graph collaborative filtering recommendation method. The method comprises: generating a first embedded representation using a spectrum GNN; forming an adaptive frequency signal extractor based on an adjacent matrix corresponding to the user-item interaction matrix; extracting a low-frequency embedded representation, an intermediate-frequency embedded representation and a high-frequency embedded representation from the first embedded representation by using the adaptive frequency signal extractor; based on the low-frequency embedding representation, the intermediate-frequency embedding representation and the high-frequency embedding representation, generating a mixed embedding representation, and performing comparative learning and recommendation prediction; and iteratively training the generative graph collaborative filtering recommendation model based on the loss function of the comparative learning and the loss function of the recommendation prediction. According to the graph collaborative filtering recommendation method based on frequency-adaptive comparative learning, signals with different frequencies can be fully utilized, and the performance of the spectrum GNN is improved.

Description

Technical Field

[0001] The present application relates to the technical field of machine learning, and in particular to a graph collaborative filtering recommendation method based on adaptive frequency contrast learning. Background Art

[0002] With the rapid development of internet technology, the problem of information overload has become increasingly prominent. Recommender systems, as a core technology for addressing this issue, have been widely used in e-commerce platforms, social media, content distribution, and other fields. Collaborative filtering (CF), a classic method in recommendation systems, analyzes user interactions with items (such as clicks, purchases, and ratings) to explore user preferences and item characteristics, thereby predicting potential interests.

[0003] However, traditional collaborative filtering methods (such as matrix factorization) face challenges such as data sparsity and cold start. In recent years, graph neural networks (GNNs) have gradually become the mainstream technology for collaborative filtering due to their ability to effectively model high-order relationships in user-item interaction graphs.

[0004] Among them, in graph neural networks, since the spatial domain GNN model is limited to a fixed polynomial basis design, the frequency response range of the filter cannot be flexibly adjusted, resulting in high-frequency signals being difficult to be effectively utilized. Therefore, spectral domain graph neural networks that can more comprehensively utilize the multi-frequency characteristics of graph signals and significantly improve recommendation performance are widely used.

[0005] However, on the one hand, the existing spectral domain GNN has extremely high computational complexity in performing eigenvalue decomposition of large-scale graph Laplacian matrices, making it difficult to apply to real-world scenarios; on the other hand, it ignores the differences in the responses of different frequency signals to contrastive learning in spectral domain GNN, resulting in poor performance of spectral domain GNN. Summary of the Invention

[0006] In order to address the deficiencies in the prior art, the purpose of this application is to provide a graph collaborative filtering recommendation method based on adaptive frequency contrast learning, which can fully utilize signals of different frequencies and improve the performance of spectral GNN.

[0007] To achieve the above objectives, the present application provides a graph collaborative filtering recommendation method based on adaptive frequency contrastive learning, comprising:

[0008] Based on the user-item interaction matrix, a spectral GNN is used to generate the first embedding representation;

[0009] Based on the adjacency matrix corresponding to the user-item interaction matrix, an adaptive frequency signal extractor is formed by approximating the preset signal extraction function;

[0010] Extracting a low-frequency embedded representation, a mid-frequency embedded representation, and a high-frequency embedded representation from the first embedded representation using the adaptive frequency signal extractor, wherein the mid-frequency embedded representation is equal to the difference between the first embedded representation and the low-frequency embedded representation and the high-frequency embedded representation;

[0011] Based on the low-frequency embedding representation, the medium-frequency embedding representation, and the high-frequency embedding representation, a hybrid embedding representation is generated, and comparative learning and recommendation prediction are performed;

[0012] Based on the contrastive learning loss function and the recommendation prediction loss function, a graph collaborative filtering recommendation model is iteratively trained and generated.

[0013] Furthermore, the spectral GNN is a JGCF spectral GNN.

[0014] Furthermore, the specific step of forming an adaptive frequency signal extractor by approximating a preset signal extraction function based on the adjacency matrix corresponding to the user-item interaction matrix adopts the following formula:

[0015]

[0016] Among them, α, β, and μ are waveform control parameters, |α| is used to control the steepness of the waveform, β is used to control the position of the waveform, and μ is used to scale the size of the extracted waveform. is the normalized adjacency matrix corresponding to the user-item interaction matrix, F() is the signal extraction function, is the adaptive frequency signal extractor, k i is a learnable weight parameter, and H represents the number of polynomial layers.

[0017] Furthermore, the method further includes training the adaptive frequency signal extractor using the following formula:

[0018]

[0019] Among them, L f is the training loss function, X is the training sample of the eigenvalue interval [-1, 1] of the adjacency matrix A.

[0020] Furthermore, the specific step of extracting the low-frequency embedded representation, the medium-frequency embedded representation, and the high-frequency embedded representation from the first embedded representation using the adaptive frequency signal extractor adopts the following formula:

[0021]

[0022] in, represents the node embedding on the original user-item graph, where E M =E J -E L -EH , E L is the low-frequency embedding representation, E M is the intermediate frequency embedding representation, E H is the high frequency embedding representation, E J is the first embedding representation, J() is the basis function of spectral GNN, An adaptive frequency signal extractor for low-frequency extraction parameters, An adaptive frequency signal extractor for IF extraction parameters, An adaptive frequency signal extractor for high frequency extraction parameters.

[0023] Furthermore, the specific steps of generating a hybrid embedding representation based on the low-frequency embedding representation, the medium-frequency embedding representation, and the high-frequency embedding representation, and performing comparative learning and recommendation prediction adopt the following formula:

[0024] E=w1E L +w2E M +w3E H ;

[0025] in represents the mixed embedding representation. w1, w2, w3 are hyperparameters that control the ratio of low-frequency, medium-frequency, and high-frequency signals;

[0026]

[0027] Among them, u is the user, U is the user set, i is the project, is the item set, v is the user in the user set U that does not belong to user u, and j is the item in the item set that does not belong to item i; Comparing loss functions for users, is the item comparison loss function, is the total contrast loss function; e u 、e i ∈E,e u is the mixed embedding representation of user u, e i is the mixed embedding representation of item i.

[0028] Furthermore, the specific steps of iteratively training and generating the graph collaborative filtering recommendation model based on the contrastive learning loss function and the recommendation prediction loss function adopt the following formula:

[0029]

[0030] Among them, e u 、e i 、e j ∈E,e u is the mixed embedding representation of user u, e i is the mixed embedding representation of item i, ej is the mixed embedding representation of item j, is the set of neighbors of user u; is the recommendation score of user u for item i, is the recommendation score of user u for item j; is the recommendation loss function; is the total contrast loss function, λ1 is the weight parameter, and λ2 is the coefficient of the regularization term.

[0031] To achieve the above-mentioned purpose, the electronic device provided by this application includes:

[0032] processor;

[0033] a memory having stored thereon one or more computer program instructions executed on the processor;

[0034] When the processor runs the computer instructions, the graph collaborative filtering recommendation method of adaptive frequency contrast learning is performed as described above.

[0035] To achieve the above objectives, the present application provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed by a processor, the steps of the graph collaborative filtering recommendation method of adaptive frequency contrast learning are executed as described above.

[0036] The present application discloses a graph collaborative filtering recommendation method with adaptive frequency contrast learning, which extracts graphic signals of different frequencies from a spectrum GNN by setting an adaptive frequency signal extractor, approximates the signal extraction function using a polynomial basis, and adaptively learns the coefficients of the basis. It also fuses the adjusted frequency signals and performs contrast learning on the adjusted frequency signals through frequency signal segmentation contrast learning, thereby making full use of different frequency signals and enhancing the recommendation performance of the GNN.

[0037] This application proposes a graph collaborative filtering recommendation method based on adaptive frequency contrast learning, which is applicable to any spectral GNN and can assist in adjusting the waveform of the original spectral GNN filter to enhance the performance of downstream tasks.

[0038] Other features and advantages of the present application will be set forth in the following description, and in part will be apparent from the description, or may be learned by practicing the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are used to provide a further understanding of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings:

[0040] Figure 1A flowchart of the graph collaborative filtering recommendation method based on adaptive frequency contrast learning in this application;

[0041] Figure 2 This is a schematic diagram of the structure of the graph collaborative filtering recommendation model of Example 1 of the present application;

[0042] Figure 3 This is a schematic diagram of the content of the data set of Example 1 of this application;

[0043] Figure 4 Schematic diagram of the recommendation performance of each recommendation model;

[0044] Figure 5 Schematic diagram of the optimal model parameters for different data sets. DETAILED DESCRIPTION

[0045] The following describes embodiments of the present application in more detail with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are for illustrative purposes only and are not intended to limit the scope of protection of the present application.

[0046] It should be understood that the various steps described in the method embodiments of the present application can be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0047] As used herein, the term "including" and its variations are open-ended, i.e., "including but not limited to." The term "based on" means "based, at least in part, on." The term "one embodiment" means "at least one embodiment," the term "another embodiment" means "at least one additional embodiment," and the term "some embodiments" means "at least some embodiments." Other terms are defined in the following description.

[0048] It should be noted that the modifications of "one" and "plurality" mentioned in this application are illustrative rather than restrictive. Those skilled in the art will understand that unless the context clearly indicates otherwise, they should be understood as "one or more." "Plurality" should be understood as two or more.

[0049] Spectral Graph Neural Networks (SGNNs) are a type of deep learning model based on spectral graph theory and graph signal processing (GSP). They aim to efficiently model graph-structured data through frequency-domain analysis. Their core concept is to map graph data into the frequency domain and, using the Graph Fourier Transform (GFT) and filter design, capture both low- and high-frequency features of graph signals, thereby solving tasks such as node classification, link prediction, and graph classification.

[0050] For example, given an undirected graph G = (V, E), let V and E be the node set and edge set of the graph. We denote the user set as The item set is denoted as J and the node set is denoted as The interaction matrix between users and items is: If the i-th user and the j-th item have an interaction, then R ij = 1. The adjacency matrix can be expressed as follows:

[0051]

[0052] Spectral GNN performs graph convolution in the spectral domain of the Laplacian matrix. The normalized Laplacian matrix can be defined as in represents the normalized adjacency matrix, Represents the degree matrix of the graph. Its eigenvalues ​​and corresponding eigenvectors are obtained by eigenvalue decomposition. where U is the eigenvector, is the corresponding eigenvalue, The eigenvalue range of the signal is [0, 2]. The graphical Fourier transform of can be defined as: The inverse transform is This transformation enables the formulation of operations such as filtering in the spectral domain. A filtering operation on a signal x can be defined as:

[0053]

[0054] f() is the graph signal filtering function. It is difficult to directly calculate f(Λ) because the eigenvalue decomposition of a large Laplacian matrix is ​​very time-consuming. Therefore, spectral GNNs often use some polynomials to approximate f(Λ):

[0055]

[0056] where θ i Typically a learnable or fixed scalar, Pi (Λ) Various forms of polynomial bases are usually used.

[0057] Graph signal, graph signal refers to the adjacency matrix The eigenvalue decomposition of obtains a series of eigenvalues ​​Λ in the range of [-1, 1]. When the eigenvalue is close to -1, it exhibits high-frequency characteristics, and when the eigenvalue is close to 1, it exhibits low-frequency characteristics.

[0058] Hereinafter, embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0059] Example 1

[0060] An embodiment of the present application provides a graph collaborative filtering recommendation method based on contrastive learning of adaptive frequency, which will be referred to below. Figure 1-Figure 5 The graph collaborative filtering recommendation method based on adaptive frequency contrast learning of the present application is described in detail.

[0061] Step S101: Generate a first embedding representation using a spectral GNN based on the user-item interaction matrix;

[0062] In this embodiment, the spectral GNN of JGCF is used to generate the first embedding representation, as follows:

[0063]

[0064] E J =[E bs ;E bp ];

[0065] Among them, E J is the first embedding representation, E bs For a band-stop filter, E bp is a bandpass filter, Represents the normalized adjacency matrix The kth Jacobi polynomial representation of the normalized adjacency matrix Λ is a graph The eigenvalues ​​of represents the node embeddings on the original user-item graph.

[0066] From the above, we can know that the adjacency matrix A can be obtained from the user-item interaction matrix, and then the normalized adjacency matrix can be obtained

[0067] In some other implementations, other spectral GNNs may also be used to obtain the first embedding representation.

[0068] Step 102: Based on the adjacency matrix corresponding to the user-item interaction matrix, an adaptive frequency signal extractor is formed by approximating a preset signal extraction function.

[0069] In this embodiment, inspired by the multiplication filter network, a signal extraction function is set. The extracted frequency interval is 1, and the interval to be shielded is set to 0, so as to multiply it with the original mixing filter to accurately extract the target frequency signal. The signal extraction function is as follows:

[0070]

[0071] Among them, α, β, and μ are waveform control parameters, |α| is used to control the steepness of the waveform, β is used to control the position of the waveform, and μ is used to scale the size of the extracted waveform. is the normalized adjacency matrix corresponding to the user-item interaction matrix, F() is the signal extraction function,

[0072] Because the exponential operation on the graph is meaningless, the signal extraction function cannot be directly applied to the GCF. The multiplication operation of the adjacency matrix corresponds to the order of the adjacency matrix, such as in Represents the 2nd-order adjacency matrix of the graph, but e*e≠e 2 Operations have no practical meaning on the graph.

[0073] Therefore, in this embodiment, an adaptive frequency signal extractor (AFSE) is obtained by using the polynomial of the adjacency matrix A to approximate the waveform of the signal extraction function, and the formula is as follows:

[0074]

[0075] Among them, F() is the signal extraction function, is the adaptive frequency signal extractor, k i is a learnable weight parameter, and H represents the number of polynomial layers.

[0076] As can be seen from the above, the image signal filtering function transforms the original image signal λ∈Λ to obtain a new image signal.

[0077] Step S103: using the adaptive frequency signal extractor to extract a low-frequency embedded representation, a medium-frequency embedded representation, and a high-frequency embedded representation from the first embedded representation;

[0078] The specific steps are as follows:

[0079]

[0080] in, represents the node embedding on the original user-item graph, where E M =EJ -E L -E H , E L is the low-frequency embedding representation, E M is the intermediate frequency embedding representation, E H is the high frequency embedding representation, E J is the first embedding representation, J() is the basis function of spectral GNN, An adaptive frequency signal extractor for low-frequency extraction parameters, An adaptive frequency signal extractor for IF extraction parameters, An adaptive frequency signal extractor for high frequency extraction parameters.

[0081] Step S104: generating a hybrid embedding representation based on the low-frequency embedding representation, the medium-frequency embedding representation, and the high-frequency embedding representation, and performing comparative learning and recommendation prediction;

[0082] In this embodiment, three hyperparameters are used to control the weights of different signals extracted from the original mixed signal. The specific formulas are as follows:

[0083] E=w1E L +w2E M +w3E H ;

[0084] in represents the mixed embedding representation. w1, w2, w3 are hyperparameters that control the ratio of low-frequency, medium-frequency, and high-frequency signals;

[0085] In this embodiment, after obtaining the hybrid embedding representation, comparative learning and recommendation prediction are also performed. The formula is as follows:

[0086]

[0087] Among them, u is the user, U is the user set, i is the project, is the item set, v is the user in the user set U that does not belong to user u, and j is the item in the item set that does not belong to item i; Comparing loss functions for users, is the item comparison loss function, is the total contrast loss function; e u 、e i ∈E,e u is the mixed embedding representation of user u, e i is the mixed embedding representation of item i.

[0088] Step S105: Iteratively training and generating a graph collaborative filtering recommendation model based on the contrastive learning loss function and the recommendation prediction loss function;

[0089] In this embodiment, Bayesian Personalized Ranking (BPR) is used as the recommendation loss function:

[0090]

[0091] Among them, e u 、e i ,j∈E,e u is the mixed embedding representation of user u, e i is the mixed embedding representation of item i, e i is the mixed embedding representation of item j, is the set of neighbors of user u; is the recommendation score of user u for item i, is the recommendation score of user u for item j; is the recommendation loss function; is the total contrast loss function, λ1 is the weight parameter, and λ2 is the coefficient of the regularization term.

[0092] Use the final loss function Training is performed to generate a collaborative filtering recommendation model (AFC-GCF).

[0093] See Figure 3 , Figure 3 This is a schematic diagram of the content of the data set of Example 1 of this application, such as Figure 3 As shown in the figure, the datasets include Gowalla, Amazon-Books, Yelp2018, and Alibaba-iFashio. Gowalla is widely used as a point-of-interest network for evaluating recommendation algorithms. Amazon-Books is a historical user rating dataset based on the Amazon online book store. Yelp2018 is a business recommendation dataset in which each business is considered an item. Alibaba-iFashio includes clicks on products by Taobao iFashion users.

[0094] In this implementation, Recall@K and NDCG@K are used as evaluation metrics, and the latest recommendation model is selected as our baseline model, where K∈[10, 20, 50]. Figure 4 , Figure 4 Schematic diagram of the recommendation performance of each recommendation model. The collaborative filtering recommendation model (AFC-GCF) of this application significantly outperforms models such as JGCF, NGCF and DGCF in various performance indicators on the Gowalla and Amazon-Books datasets, with an improvement of 1%-15% compared to these models.

[0095] For example, see Figure 5 , Figure 5Schematic diagram of the optimal model parameters for different data sets. In this embodiment, the collaborative filtering recommendation model is trained to generate the optimal values ​​of parameters such as graph convolution model depth, initialization embedding dimension, batch size, learning rate, temperature hyperparameter, λ1, λ2 and H (order of polynomial function) on different data sets.

[0096] Example 2

[0097] This embodiment also provides an electronic device comprising a processor and a memory. The memory is configured to store non-transitory computer-readable instructions. The processor is configured to execute the non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by the processor, one or more steps of the graph collaborative filtering recommendation method using adaptive frequency contrastive learning are performed as described above. The memory and the processor may be interconnected via a bus system and / or other connection mechanisms.

[0098] For example, the processor can be a central processing unit (CPU), a digital signal processor (DSP), or other forms of processing units with data processing capabilities and / or program execution capabilities, such as a field programmable gate array (FPGA); for example, the central processing unit (CPU) can be an X86 or ARM architecture, etc.

[0099] For example, the memory may include any combination of one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, a flash memory, etc. One or more computer program modules may be stored on the computer-readable storage medium, and the processor may execute one or more computer program modules to implement various functions of the electronic device. Various applications and various data, as well as various data used and / or generated by the application, may also be stored in the computer-readable storage medium.

[0100] It should be noted that, in the embodiments of the present application, the specific functions and technical effects of the electronic device can refer to the description of the graph collaborative filtering recommendation method for contrastive learning of adaptive frequency in the above text, which will not be repeated here.

[0101] Example 3

[0102] This embodiment further provides a computer-readable storage medium for storing non-transitory computer-readable instructions. For example, when the non-transitory computer-readable instructions are executed by a computer, one or more steps of the graph collaborative filtering recommendation method based on adaptive frequency contrastive learning described above can be performed.

[0103] For example, the storage medium can be applied to the above-mentioned electronic device. For example, the storage medium can be the memory in the electronic device of Example 2. For example, the relevant description of the storage medium can refer to the corresponding description of the memory in the electronic device of Example 2, and will not be repeated here.

[0104] It should be noted that the storage medium (computer-readable medium) mentioned above in the present application may be a computer-readable signal medium or a non-transitory computer-readable storage medium or any combination of the above two. The non-transitory computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of non-transitory computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0105] In this application, a non-transitory 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 this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a non-transitory computer-readable storage medium that can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0106] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0107] Computer program code for carrying out the operations of the present application may be written in one or more programming languages, or a combination thereof, including, but not limited to, object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may execute entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server.

[0108] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code includes one or more executable instructions for realizing the prescribed logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented using a dedicated hardware-based system that performs the prescribed function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0109] The units involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of a unit does not, in some cases, constitute a limitation on the unit itself.

[0110] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0111] The above description is only a partial embodiment of the present application and an illustration of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in this application is not limited to the technical solutions formed by the specific combination of the above technical features, but also includes other technical solutions formed by any combination of the above technical features or their equivalents without departing from the above disclosed concepts. For example, the above features can be replaced with (but not limited to) technical features with similar functions disclosed in this application.

[0112] In addition, although adopting specific order to describe each operation, this should not be interpreted as requiring these operations to be performed in the specific order shown or in sequential order. Under certain environment, multitasking and parallel processing may be advantageous. Similarly, although comprising some specific implementation details in the above discussion, these should not be interpreted as limiting the scope of the application. Some features described in the context of separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment also can be implemented in multiple embodiments individually or in the mode of any suitable subcombination.

[0113] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

1. A graph collaborative filtering recommendation method based on adaptive frequency contrastive learning, comprising: Based on the user-item interaction matrix, a spectral GNN is used to generate the first embedding representation; Based on the adjacency matrix corresponding to the user-item interaction matrix, an adaptive frequency signal extractor is formed by approximating the preset signal extraction function; Extracting a low-frequency embedded representation, a mid-frequency embedded representation, and a high-frequency embedded representation from the first embedded representation using the adaptive frequency signal extractor, wherein the mid-frequency embedded representation is equal to the difference between the first embedded representation and the low-frequency embedded representation and the high-frequency embedded representation; Based on the low-frequency embedding representation, the medium-frequency embedding representation, and the high-frequency embedding representation, a hybrid embedding representation is generated, and comparative learning and recommendation prediction are performed; Based on the contrastive learning loss function and the recommendation prediction loss function, a graph collaborative filtering recommendation model is iteratively trained and generated.

2. The graph collaborative filtering recommendation method based on adaptive frequency contrast learning according to claim 1 is characterized in that: The spectral GNN is a JGCF spectral GNN.

3. The graph collaborative filtering recommendation method based on adaptive frequency contrast learning according to claim 2 is characterized in that: The specific step of forming an adaptive frequency signal extractor by approximating a preset signal extraction function based on the adjacency matrix corresponding to the user-item interaction matrix adopts the following formula: Among them, α, β, and μ are waveform control parameters, |α| is used to control the steepness of the waveform, β is used to control the position of the waveform, and μ is used to scale the size of the extracted waveform. is the normalized adjacency matrix corresponding to the user-item interaction matrix, F() is the signal extraction function, is the adaptive frequency signal extractor, k i is a learnable weight parameter, and H represents the number of polynomial layers.

4. The graph collaborative filtering recommendation method based on adaptive frequency contrast learning according to claim 3 is characterized in that: It also includes training the adaptive frequency signal extractor using the following formula: Among them, L f is the training loss function, X is the training sample of the eigenvalue interval [-1, 1] of the adjacency matrix A.

5. The graph collaborative filtering recommendation method based on adaptive frequency contrast learning according to claim 4 is characterized in that: The specific steps of extracting the low-frequency embedded representation, the medium-frequency embedded representation, and the high-frequency embedded representation from the first embedded representation using the adaptive frequency signal extractor are as follows: in, represents the node embedding on the original user-item graph, where E M =E J -E L -E H , E L is the low-frequency embedding representation, E M is the intermediate frequency embedding representation, E H is the high frequency embedding representation, E J is the first embedding representation, J() is the basis function of spectral GNN, An adaptive frequency signal extractor for low-frequency extraction parameters, An adaptive frequency signal extractor for IF extraction parameters, An adaptive frequency signal extractor for high frequency extraction parameters.

6. The graph collaborative filtering recommendation method based on adaptive frequency contrast learning according to claim 5, characterized in that: The specific steps of generating a hybrid embedding representation based on the low-frequency embedding representation, the medium-frequency embedding representation, and the high-frequency embedding representation, and performing comparative learning and recommendation prediction are as follows: E=w1E L +w2E M +w3E H ; in represents the mixed embedding representation, w1, w2, w3 are hyperparameters that control the proportion of low-frequency, medium-frequency and high-frequency signals; Among them, u is the user, U is the user set, i is the project, is the item set, v is the user in the user set U that does not belong to user u, and j is the item in the item set that does not belong to item i; Comparing loss functions for users, is the item comparison loss function, is the total contrast loss function; e u 、e i ∈E,e u is the mixed embedding representation of user u, e i is the mixed embedding representation of item i.

7. The graph collaborative filtering recommendation method based on adaptive frequency contrast learning according to claim 6, characterized in that: The specific steps of iteratively training and generating the graph collaborative filtering recommendation model based on the contrastive learning loss function and the recommendation prediction loss function are as follows: Among them, e u 、e i ,j∈E,e u is the mixed embedding representation of user u, e i is the mixed embedding representation of item i, e i is the mixed embedding representation of item j, is the set of neighbors of user u; is the recommendation score of user u for item i, is the recommendation score of user u for item j; is the recommendation loss function; is the total contrast loss function, λ1 is the weight parameter, and λ2 is the coefficient of the regularization term.

8. An electronic device, characterized in that: include: processor; a memory having stored thereon one or more computer program instructions executed on the processor; When the processor runs the computer instructions, it executes the graph collaborative filtering recommendation method of adaptive frequency contrast learning according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that Computer instructions are stored thereon, and when the computer instructions are executed by the processor, the steps of the graph collaborative filtering recommendation method of adaptive frequency contrast learning according to any one of claims 1 to 7 are executed.

Images