Recommendation method and device based on federal learning, program product and electronic equipment

Abstract

The invention provides a federal learning-based recommendation method and device, a program product and electronic equipment, and relates to the technical field of computers. The method is applied to a distributed recommendation system, the distributed recommendation system comprises a central service device and a plurality of local service devices, and the method comprises the following steps: the central service device receives respective corresponding second data sets sent by the plurality of local service devices, the second data set is obtained by performing recall processing on data in the corresponding first data set by the local service equipment by using a recall model; the central service equipment performs preliminary screening on the obtained second data set to obtain a third data set corresponding to the local service equipment, and sends the third data set to the corresponding local service equipment; and the central service equipment performs clustering sorting processing on the obtained fourth data set according to the associated information of the to-be-recommended user to obtain target information recommended to the to-be-recommended user. The recommendation efficiency and accuracy of the recommendation system are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a recommendation method, device, program product and electronic device based on federated learning. Background Art

[0002] At present, as one of the most widely used fields of artificial intelligence, recommendation systems can give corresponding recommendations based on user preferences, help users make choices, and improve user experience. Therefore, how to improve the recommendation efficiency and accuracy of recommendation systems has become an urgent problem to be solved. Summary of the invention

[0003] The present disclosure provides a recommendation method based on federated learning, a recommendation device based on federated learning, a program product and an electronic device, so as to at least solve the problems of low recommendation efficiency and poor accuracy of the recommendation system to a certain extent.

[0004] According to a first aspect of the present disclosure, a recommendation method based on federated learning is provided, which is applied to a distributed recommendation system, wherein the distributed recommendation system includes a central service device and multiple local service devices, wherein a recall model and a refined ranking model are deployed in the local service devices; the method includes:

[0005] The central service device receives second data sets corresponding to each other sent by a plurality of local service devices, where the second data sets are obtained by the local service devices recalling data in the corresponding first data sets using the recall model;

[0006] The central service device preliminarily screens the obtained second data set to obtain a third data set corresponding to the local service device, and sends the third data set to the corresponding local service device respectively;

[0007] The central service device clusters and sorts the fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device sorting the data in the corresponding third data set using the refined sorting model.

[0008] In one implementation, the central service device performs clustering and sorting processing on the obtained fourth data set according to the associated information of the user to be recommended to obtain the target information recommended to the user to be recommended, including:

[0009] The central service device determines the target comparison object according to the associated information of the user to be recommended;

[0010] The central service device determines the correlation value between the data to be compared in each of the fourth data sets and the target comparison object, and determines the recommended candidate set to which each of the data to be compared belongs according to the correlation value corresponding to each of the data to be compared;

[0011] The central service device determines the target information to be recommended to the to-be-recommended user according to the ranking information of each recommendation candidate set.

[0012] In one implementation, the central service device determines the target comparison object according to the associated information of the user to be recommended, including:

[0013] The central service device determines, based on the association information of the user to be recommended, the keywords of interest of the user to be recommended to the objects in the first data set;

[0014] The central service device will generate a keyword portrait based on the interest keywords as the target comparison object.

[0015] In one implementation, the central service device determines the recommendation candidate set to which each of the to-be-compared data belongs according to the relevance value corresponding to each of the to-be-compared data, including:

[0016] The central service device determines the relevance ranges corresponding to the plurality of recommendation candidate sets respectively;

[0017] When the central service device determines that the relevance value corresponding to the first data to be compared is within the relevance range corresponding to the first recommendation candidate set, it is determined that the first data to be compared belongs to the first recommendation candidate set; the first data to be compared belongs to any of the fourth data sets.

[0018] In one implementation, after the central service device obtains the target information recommended to the user to be recommended, the method further includes:

[0019] The central service device determines a target display format for the data in the target information according to the historical display information in the associated information of the user to be recommended;

[0020] The central service device sends the target information and the target display form to the terminal device corresponding to the to-be-recommended user, so that the terminal device displays the data in the target information to the to-be-recommended user in the target display form.

[0021] According to a second aspect of the present disclosure, a recommendation method based on federated learning is provided, which is applied to a distributed recommendation system, wherein the distributed recommendation system includes a central service device and multiple local service devices, wherein a recall model and a refined ranking model are deployed in the local service devices; the method includes:

[0022] The local service device uses the recall model to perform recall processing on the data in the corresponding first data set to obtain the corresponding second data set, and sends the second data set to the central service device;

[0023] The local service device receives a corresponding third data set sent by the central service device; the third data set is obtained by the central service device performing a preliminary screening on the corresponding second data set;

[0024] The local service device uses the refined sorting model to sort the data in the corresponding third data set to obtain the corresponding fourth data set, and sends the fourth data set to the central service device, so that the central service device clusters and sorts the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended.

[0025] In one implementation, the local service device uses the recall model to perform recall processing on each data in the corresponding first data set to obtain the corresponding second data set, including:

[0026] The local service device calculates the recall score of each data in the first data set by using the recall model, and obtains a processed data set including the recall score;

[0027] The local service device rearranges the data in the processed data set in a key-value pair manner to obtain a second data set.

[0028] In one embodiment, the recall model is trained based on the following method:

[0029] The local service device calculates the gradient tensor of the untrained recall model using the loss function, encrypts the gradient tensor using an encryption algorithm to obtain a ciphertext, and sends the ciphertext to the central service device;

[0030] The local service device calculates the decrypted ciphertext result and updates the untrained recall model by back propagation until the updated recall models of each local service device converge, so as to obtain the recall model; wherein the ciphertext calculation result is obtained by the central service device performing linear mean calculation on each ciphertext received.

[0031] According to a third aspect of the present disclosure, a recommendation device based on federated learning is provided, which is applied to a distributed recommendation system, wherein the distributed recommendation system includes a central service device and multiple local service devices, wherein a recall model and a refined ranking model are deployed in the local service devices; the device includes:

[0032] A receiving unit, configured to receive second data sets corresponding to each other and sent by a plurality of local service devices, wherein the second data sets are obtained by the local service devices recalling data in the corresponding first data sets using the recall model;

[0033] a primary screening unit, configured to perform a primary screening on the obtained second data set, obtain a third data set corresponding to the local service device, and send the third data set to the corresponding local service device respectively;

[0034] The recommendation unit is used to cluster and sort the obtained fourth data set according to the associated information of the user to be recommended, so as to obtain target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined ranking model to sort the data in the corresponding third data set.

[0035] In one embodiment, the recommendation unit is used to:

[0036] Determine the target comparison object according to the associated information of the user to be recommended;

[0037] Determine the correlation value between the data to be compared in each of the fourth data sets and the target comparison object, and determine the recommended candidate set to which each of the data to be compared belongs according to the correlation value corresponding to each of the data to be compared;

[0038] According to the ranking information of each recommendation candidate set, the target information recommended to the to-be-recommended user is determined.

[0039] In one embodiment, the recommendation unit is used to:

[0040] Determining, based on the association information of the user to be recommended, keywords of interest to the user to be recommended for the objects in the first data set;

[0041] A keyword portrait will be generated based on the interest keywords as the target comparison object.

[0042] In one embodiment, the recommendation unit is used to:

[0043] Determine the relevance ranges corresponding to the multiple recommendation candidate sets;

[0044] When it is determined that the correlation value corresponding to the first data to be compared is within the correlation range corresponding to the first recommendation candidate set, it is determined that the first data to be compared belongs to the first recommendation candidate set; the first data to be compared belongs to any of the fourth data sets.

[0045] In one embodiment, the device further comprises a display unit for:

[0046] Determining a target display format for the data in the target information according to the historical display information in the associated information of the user to be recommended;

[0047] The target information and the target display form are sent to a terminal device corresponding to the to-be-recommended user, so that the terminal device displays the data in the target information to the to-be-recommended user in the target display form.

[0048] According to a fourth aspect of the present disclosure, a recommendation device based on federated learning is provided, which is applied to a distributed recommendation system, wherein the distributed recommendation system includes a central service device and multiple local service devices, wherein a recall model and a refined ranking model are deployed in the local service devices; the device includes:

[0049] A recall unit, configured to use the recall model to perform recall processing on the data in the corresponding first data set, obtain the corresponding second data set, and send the second data set to the central service device;

[0050] A receiving unit, configured to receive a corresponding third data set sent by the central service device; the third data set is obtained by the central service device performing a preliminary screening on the corresponding second data set;

[0051] The processing unit is used to use the refined ranking model to sort the data in the corresponding third data set to obtain the corresponding fourth data set, and send the fourth data set to the central service device, so that the central service device can cluster and sort the obtained fourth data set according to the associated information of the user to be recommended, and obtain the target information recommended to the user to be recommended.

[0052] In one embodiment, the recall unit is used to:

[0053] Using the recall model, calculate the recall score of each data in the first data set to obtain a processed data set including the recall score;

[0054] The data in the processed data set is rearranged in a key-value pair manner to obtain a second data set.

[0055] In one embodiment, the apparatus comprises a training unit for:

[0056] Calculating the gradient tensor of the untrained recall model using the loss function, encrypting the gradient tensor using an encryption algorithm to obtain a ciphertext, and sending the ciphertext to the central service device;

[0057] According to the decrypted ciphertext calculation result, the untrained recall model is updated by back propagation until the updated recall models of each local service device converge, so as to obtain the recall model; wherein the ciphertext calculation result is obtained by the central service device performing linear mean calculation on each ciphertext received.

[0058] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method of the first aspect and the second aspect described above and possible implementation methods thereof are implemented.

[0059] According to a sixth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the methods of the first and second aspects and possible implementations thereof by executing the executable instructions.

[0060] The technical solution disclosed in this disclosure has the following beneficial effects:

[0061] In the disclosed embodiment, since the distributed recommendation system includes a central service device and multiple local service devices, and the recall model and the refined ranking model are deployed in the local service device. Therefore, the central service device can receive the second data sets corresponding to each other sent by multiple local service devices, and the second data set is obtained by the local service device using the recall model to recall the data in the corresponding first data set; that is, in the disclosed embodiment, federated learning is combined with the recall module in the recommendation system to achieve data screening without data going out of the domain during the recall stage. And, in order to reduce the amount of data processing, the central service device can perform a preliminary screening on the obtained second data set, obtain the third data set corresponding to the local service device, and then send the third data set to the corresponding local service device respectively. In other words, the refined ranking model is deployed in the local service device for sorting processing, which speeds up the recommendation sorting processing of the third data set, shortens the processing time of the entire recommendation system, and improves efficiency. Furthermore, in order to improve the accuracy of recommendation, the central service device also performs cluster sorting processing on the fourth data set obtained according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined sorting model to sort the data in the corresponding third data set. It can be seen that the embodiment of the present disclosure adds auxiliary sorting information on the basis of the sorting of the refined sorting module, thereby improving the recommendation accuracy of the recommendation system.

[0062] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or be understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments of the present disclosure will be briefly introduced below. Obviously, the drawings introduced below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 A schematic diagram of an application scenario in this exemplary embodiment is shown;

[0065] Figure 2 A flowchart showing a recommendation method based on federated learning in this exemplary embodiment;

[0066] Figure 3 A flowchart showing another recommendation method based on federated learning in this exemplary embodiment;

[0067] Figure 4 A flowchart showing another recommendation method based on federated learning in this exemplary embodiment;

[0068] Figure 5 A schematic diagram showing a method of determining target information in this exemplary embodiment;

[0069] Figure 6 A schematic diagram showing the structure of a recommendation device based on federated learning in this exemplary embodiment is shown;

[0070] Figure 7 A schematic diagram showing the structure of a recommendation device based on federated learning in this exemplary embodiment is shown;

[0071] Figure 8 A schematic structural diagram of an electronic device in this exemplary embodiment is shown. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure. In the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be arbitrarily combined with each other. In addition, although the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in an order different from that here.

[0073] The term "comprising" and any variations thereof in the specification and claims of the present disclosure are intended to cover non-exclusive protection. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0074] In the embodiments of the present disclosure, one or more, "many" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or plural.

[0075] It should be noted that the terms "first", "second", "third", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order, sequence, size, and priority. For example, the first recommendation bucket and the second recommendation bucket in the embodiment of the present disclosure are only used to distinguish different recommendation buckets. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0076] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, which are schematic diagrams of the present disclosure and are not necessarily drawn to scale. Some of the block diagrams shown in the accompanying drawings may be functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in hardware modules or integrated circuits, or in networks, processors or microcontrollers. The implementation method can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein. The features, structures or characteristics described in the present disclosure may be combined in one or more implementations in any suitable manner. In the following description, many specific details are provided to provide a full description of the implementation methods of the present disclosure. However, those skilled in the art should be aware that one or more specific details may be omitted when implementing the technical solution of the present disclosure, or one or more specific details may be replaced by other methods, components, devices, steps, etc.

[0077] It should be noted that in the embodiments of the present disclosure, some existing solutions in the industry such as certain software, components, models, etc. may be mentioned, which should be considered as exemplary, and their purpose is only to illustrate the feasibility of implementing the technical solution of the present disclosure, but it does not mean that the applicant has or will necessarily use the solution. In the technical solution of the present disclosure, the collection, dissemination, and use of each data set, such as the data in the first data set, are in compliance with relevant national laws and regulations.

[0078] In order to facilitate understanding of the technical solution provided by the embodiments of the present disclosure, the key terms are first introduced below:

[0079] 1. Federated learning: Federated learning is a machine learning framework that can effectively help multiple institutions use data and conduct machine learning modeling while meeting the requirements of user privacy protection, data security, and government regulations. For security reasons, encryption algorithms are mostly used in federated learning to ensure the security of data transmission. The commonly used encryption algorithm is semi-homomorphic encryption.

[0080] 2. Semi-homomorphic encryption: Homomorphic encryption that only supports one of multiplication or addition. Homomorphic encryption means that the ciphertext can be directly calculated, and the result of the ciphertext calculation is the same as the result of the direct calculation of the plaintext after decryption.

[0081] 3. Clustering: Clustering is to find the intrinsic structure between data and organize all data samples into some similar clusters according to specific standards (such as distance criteria, similarity coefficient); so that the data samples in the same cluster are as similar as possible, and the samples in different clusters are as different as possible.

[0082] As one of the most widely used fields of artificial intelligence, the recommendation system can give corresponding recommendations based on the user's preferences to help the user make choices. In addition, with the continuous expansion of data scale and the increasing number of data types, information such as commodities or virtual products available for users to choose is gradually distributed in more and more companies and institutions, forming data islands. The proposal of federated learning is to allow companies and institutions to collaborate without exchanging data under the premise of meeting privacy protection and data security, thereby breaking down the barriers between data islands.

[0083] In related technologies, in order to improve the diversity and security of recommendation systems, researchers apply federated learning to recommendation systems so as to achieve more accurate recommendations without leaving the domain. However, most studies are limited to combining the algorithm of the fine ranking module in the recommendation system with the federated learning framework. However, due to the large number of model parameters used in the fine ranking module, the algorithm of the fine ranking module is executed on a single service device, which will cause a sharp drop in the inference speed after combining with the federated learning framework, reducing the recommendation efficiency of the recommendation system. In addition, after the combination of federated learning and the recommendation system, since the data is distributed in various data islands during the federated learning calculation, there are differences in the distribution of the data extracted in each round of model training, which limits the accuracy of the model's recommendations.

[0084] In view of one or more of the above problems, an exemplary embodiment of the present disclosure provides a recommendation method based on federated learning, in which a distributed recommendation system includes a central service device and multiple local service devices, and a recall model and a refined ranking model are deployed in the local service device. Therefore, the central service device can receive the second data sets corresponding to each other sent by multiple local service devices, and the second data set is obtained by the local service device using the recall model to recall the data in the corresponding first data set; that is, in the embodiment of the present disclosure, federated learning is combined with the recall module in the recommendation system to implement data screening without data out of the domain during the recall stage. And, in order to reduce the amount of data processing, the central service device can preliminarily screen the obtained second data set, obtain the third data set corresponding to the local service device, and then send the third data set to the corresponding local service device respectively. In other words, the refined ranking model is deployed in the local service device for sorting processing, which speeds up the recommendation sorting processing of the third data set, shortens the processing time of the entire recommendation system, and improves efficiency. Furthermore, in order to improve the accuracy of recommendation, the central service device also performs cluster sorting processing on the fourth data set obtained according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined sorting model to sort the data in the corresponding third data set. It can be seen that the embodiment of the present disclosure adds auxiliary sorting information on the basis of the sorting of the refined sorting module, thereby improving the recommendation accuracy of the recommendation system. In order to better understand the technical solution provided by the embodiment of the present disclosure, the application scenarios to which the technical solution provided by the embodiment of the present disclosure is applicable are briefly introduced below. It should be noted that the application scenarios introduced below are only used to illustrate the embodiment of the present disclosure and are not limited. In specific implementation, the technical solution provided by the embodiment of the present disclosure can be flexibly applied according to actual needs.

[0085] In the embodiments of the present disclosure, the recommendation method based on federated learning can be applied to various business scenarios that require recommendations, such as commodities, videos (such as short videos or dramas), songs, games, etc. For example, it can be applied to business scenarios for short video recommendations, which is not limited in the embodiments of the present disclosure.

[0086] See also Figure 1 As shown, Figure 1This is an application scenario to which the technical solution of the embodiment of the present disclosure can be applied. In the scenario diagram, multiple local service devices 101, a central service device 102 and a terminal device 103 are included. Among them, multiple local service devices 101 and the central service device 102 belong to a distributed recommendation system. The terminal device 103 and the central service device 102, the local service device 101 and the central service device 102, and each local service device 101 can be directly or indirectly connected through one or more networks 104.

[0087] In the disclosed embodiment, the user can determine to trigger a recommendation request based on the terminal device 103, so that the central service device 102 can receive the trigger request and link the local service device 101 of the distributed recommendation system to process it. Optionally, the recall model deployed in each local service device 101 performs a recall process on each data in the corresponding first data set to obtain the corresponding second data set, and sends the second data set to the central service device 102; the central service device 102 performs a preliminary screening on the obtained second data set to obtain the third data set corresponding to each local service device 101, and sends the third data set to the corresponding local service device 101 respectively; the refined sorting model deployed in each local service device 101 performs a sorting process on each data in the corresponding third data set to obtain the corresponding fourth data set, and sends the fourth data set to the central service device 102; then the central service device 102 performs a cluster sorting process on the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended.

[0088] in, Figure 1 The terminal device 103 may be a mobile phone, a tablet computer (PAD), a personal computer (PC), a smart TV, a smart watch, a smart speaker, a smart car device, and a wearable device, but is not limited thereto. These devices may have the function of supporting triggering a recommendation request.

[0089] as well as, Figure 1 The local service device 101 and the central service device 102 may be independent physical servers, or a server cluster or distributed system composed of multiple physical servers. They may also be cloud servers that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), as well as big data and artificial intelligence platforms, but are not limited to these.

[0090] Of course, the method provided in the embodiment of the present disclosure is not limited to Figure 1The application scenarios shown may also be used in other possible application scenarios, and the embodiments of the present disclosure are not limited thereto.

[0091] To further illustrate the technical solution provided by the embodiments of the present disclosure, this is described in detail below in conjunction with the accompanying drawings and specific implementation methods. Although the embodiments of the present disclosure provide method operation steps as shown in the following embodiments or drawings, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiments of the present disclosure. The method may be executed in the order of the method shown in the embodiments or drawings or in parallel during the actual processing process or when the device is executed.

[0092] See also Figure 2 , Figure 2 FIG. 1 is a flow chart of a recommendation method based on federated learning in an embodiment of the present disclosure. The flow of the method can be executed by a distributed recommendation system, for example, the distributed recommendation system includes Figure 1 The local service device 101 and the central service device 102 in the embodiment of the present invention are as follows:

[0093] Step 201: The local service device uses the recall model to perform recall processing on the data in the corresponding first data set, obtains the corresponding second data set, and sends the second data set to the central service device.

[0094] Step 202: The central service device receives the corresponding second data set sent by the local service device, performs a preliminary screening on the obtained second data set, obtains a third data set corresponding to the local service device, and sends the third data set to the corresponding local service device;

[0095] Step 203: the local service device receives the corresponding third data set sent by the central service device, sorts the data in the corresponding third data set using the refined sorting model, obtains the corresponding fourth data set, and sends the fourth data set to the central service device;

[0096] Step 204: The central service device receives the corresponding fourth data set sent by the local service device.

[0097] Step 205: The central service device performs clustering and sorting processing on the obtained fourth data set according to the associated information of the user to be recommended, and obtains target information recommended to the user to be recommended.

[0098] In the embodiment of the present disclosure, since the distributed recommendation system includes a central service device and multiple local service devices, and the recall model and the refined ranking model are deployed in the local service device. Therefore, the central service device can receive the corresponding second data sets sent by multiple local service devices, and the second data set is obtained by the local service device using the recall model to recall the data in the corresponding first data set; that is, in the embodiment of the present disclosure, federated learning is combined with the recall module in the recommendation system to implement data screening in the recall stage without data out of the domain. And, in order to reduce the amount of data processing, the central service device can preliminarily screen the obtained second data set, obtain the third data set corresponding to the local service device, and then send the third data set to the corresponding local service device respectively. In other words, the refined ranking model is deployed in the local service device for sorting processing, which speeds up the recommendation sorting processing of the third data set and shortens the processing time of the entire recommendation system. Further, in order to improve the recommendation accuracy, the central service device also performs cluster sorting processing on the obtained fourth data set in combination with the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined ranking model to sort the data in the corresponding third data set. It can be seen that the embodiment of the present disclosure adds auxiliary sorting information on the basis of the sorting by the refined sorting module, thereby improving the recommendation accuracy of the recommendation system.

[0099] Combine the following Figure 3 This paper introduces the recommendation method based on federated learning on the local service device side.

[0100] Step 301: The local service device uses the recall model to perform recall processing on the data in the corresponding first data set, obtains the corresponding second data set, and sends the second data set to the central service device.

[0101] In this exemplary embodiment, the local service device can use the recall model to calculate the recall score of each data in the first data set, obtain a processed data set including the recall score, and re-adjust the data in the processed data set in the form of key-value pairs to obtain a second data set. The recall model is used to screen out candidate data of the thousand level from the candidate data of the million level or the ten million level stored in the local service device, and the recall model is, for example, a double tower model.

[0102] For example, the local service device obtains a short video data set, which can be a local candidate set of tens of millions or millions, and this data set is the first data set. The local device calculates the recall score for each short video in the first data set through the recall model, and semi-homomorphically encrypts the second data set including the "short video-score" key-value pair, and transmits it to the central service device.

[0103] In this exemplary embodiment, the recall model is trained based on the following method: the local service device uses a loss function to calculate the gradient tensor of an untrained recall model, and encrypts the gradient tensor using an encryption algorithm to obtain a ciphertext, and sends the ciphertext to the central service device; the local service device calculates the result of the decrypted ciphertext and updates the untrained recall model by back propagation until the updated recall model of the local service device converges to obtain the recall model; wherein the ciphertext calculation result is obtained by the central service device performing a linear mean calculation on each ciphertext received.

[0104] It can be seen that in this exemplary embodiment, each local service device is deployed with its own corresponding recall model. In other words, the recall models deployed on each local service device in the distributed recommendation system may be the same or different.

[0105] Step 302: The local service device receives the corresponding third data set sent by the central service device; the third data set is obtained by the central service device through preliminary screening of the corresponding second data set;

[0106] Step 303: The local service device uses the refined sorting model to sort the data in the corresponding third data set to obtain the corresponding fourth data set, and sends the fourth data set to the central service device, so that the central service device clusters and sorts the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended.

[0107] Optionally, the refined ranking models are, for example, factorization machines (FM), field-aware factorization machines (FFM), deep cross networks (DCN), deep factorization machines DeepFM and deep learning recommendation system XdeepFM, etc. The local service device can select a refined ranking model suitable for local data based on the amount of local data, i.e. the amount of data corresponding to the third data set, the type of data features, etc.

[0108] It can be seen that in this exemplary embodiment, the recall model and the refined ranking model are deployed in the local service device, that is, the data screening and preliminary sorting processes are set in multiple local service devices, thereby reducing the processing volume of the central service device and further improving the recommendation efficiency of the recommendation system.

[0109] Combine the following Figure 4 This paper introduces the recommendation method based on federated learning on the central service device side.

[0110] Step 401: The central service device receives second data sets corresponding to each other sent by a plurality of local service devices, where the second data sets are obtained by the local service devices recalling data in the corresponding first data sets using a recall model.

[0111] In this exemplary embodiment, the method for obtaining the second data set can refer to the above-mentioned step 301, which will not be described in detail here.

[0112] Step 402: The central service device preliminarily screens the obtained second data set to obtain the third data set corresponding to the local service device, and sends the third data set to the corresponding local service device.

[0113] For example, if the second data set is a data set including a key-value pair of "short video-score", the central service device can filter the first N data from high to low according to the score, where N is a positive integer, to form a third data set. That is, the third data set is the short video data set after the initial screening. It should be noted that since the second data set corresponding to each local service device is different, the value of N may also be different.

[0114] Step 403: The central service device performs clustering and sorting processing on the fourth data set obtained according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined sorting model to sort the data in the corresponding third data set.

[0115] In this exemplary embodiment, the central service device may determine the target comparison object based on the association information of the user to be recommended. Optionally, the central service device determines the interest keywords of the user to be recommended for the objects in the first data set based on the association information of the user to be recommended; the central service device generates a keyword portrait as the target comparison object based on the interest keywords. Among them, the association information of the user to be recommended may be carried in the recommendation request triggered by the user, or may be determined based on the information expansion carried in the recommendation request triggered by the user, which is not limited in this exemplary embodiment. In addition, the central service device may generate a keyword portrait based on a language model, such as a natural language processing model (BidirectionalEncoder Representations from Transformers, Bert), a generative pre-trained Transformer model (Generative Pre-trained Transformer, GPT), and other models, combined with interest keywords, and use the keyword portrait as the target comparison object.

[0116] In this exemplary embodiment, the central service device determines the correlation value between the data to be compared and the target comparison object in each fourth data set, and determines the recommended candidate set to which each data to be compared belongs according to the correlation value corresponding to each data to be compared.

[0117] Optionally, the central service device determines the relevance ranges corresponding to multiple recommendation candidate sets respectively; when the central service device determines that the relevance value corresponding to the first data to be compared is within the relevance range corresponding to the first recommendation candidate set, it is determined that the first data to be compared belongs to the first recommendation candidate set; the first data to be compared belongs to any fourth data set.

[0118] For example, see Figure 5 , when it is determined that any data in the fourth data set, for example, the first data to be compared has a correlation value of 0.82, and the correlation range of the first candidate recommendation set is determined to be 0.80 to 0.99, then it is determined that the first data to be compared belongs to the first recommendation candidate set. Figure 5 When it is determined that the correlation value corresponding to the second data to be compared in another fourth data set is 0.73, and it is determined that the correlation range corresponding to the second candidate recommendation set is 0.70-0.79, it is determined that the second data to be compared belongs to the second recommendation candidate set.

[0119] In this exemplary embodiment, after all recommendation candidate sets are obtained, the central service device may determine the target information to be recommended to the user to be recommended based on the ranking information of each recommendation candidate set.

[0120] Optionally, the central service device may use the recommendation candidate set with the highest level of sorting information as the target information recommended to the user to be recommended, or may use the recommendation candidate set with the lowest level of sorting information as the target information recommended to the user to be recommended. Of course, the recommendation candidate set corresponding to the sorting information may also be freely selected as the target information recommended to the user to be recommended, and this exemplary embodiment does not limit this.

[0121] In one embodiment, after the central service device obtains the target information recommended to the user to be recommended, the central service device can also determine the target display form of the data in the target information based on the historical display information in the associated information of the user to be recommended; the central service device sends the target information and the target display form to the terminal device corresponding to the user to be recommended, so that the terminal device displays the data in the target information to the user to be recommended in the target display form.

[0122] Optionally, if the data is a short video, the target display format is to arrange the short videos in 3 rows and 8 columns and only display the content of the short video cover. The target display format can also be to display the short videos in chronological order of their release time. Of course, there can be other display formats, which are not limited to this in this exemplary embodiment.

[0123] The exemplary embodiment of the present disclosure also provides a recommendation device based on federated learning. Figure 6 As shown, the recommendation device 600 based on federated learning includes the following program modules:

[0124] A recall unit 601 is used to use the recall model to perform recall processing on the data in the corresponding first data set, obtain the corresponding second data set, and send the second data set to the central service device;

[0125] The receiving unit 602 is used to receive a corresponding third data set sent by the central service device; the third data set is obtained by the central service device performing a preliminary screening on the corresponding second data set;

[0126] Processing unit 603 is used to use the refined ranking model to sort the data in the corresponding third data set to obtain the corresponding fourth data set, and send the fourth data set to the central service device, so that the central service device can cluster and sort the obtained fourth data set according to the associated information of the user to be recommended, and obtain the target information recommended to the user to be recommended.

[0127] In one implementation, the recall unit 601 is used to:

[0128] Using the recall model, calculate the recall score of each data in the first data set to obtain a processed data set including the recall score;

[0129] The data in the processed data set is rearranged in a key-value pair manner to obtain a second data set.

[0130] In one embodiment, the apparatus comprises a training unit for:

[0131] Calculating the gradient tensor of the untrained recall model using the loss function, encrypting the gradient tensor using an encryption algorithm to obtain a ciphertext, and sending the ciphertext to the central service device;

[0132] According to the decrypted ciphertext calculation result, the untrained recall model is updated by back propagation until the updated recall models of each local service device converge, so as to obtain the recall model; wherein the ciphertext calculation result is obtained by the central service device performing linear mean calculation on each ciphertext received.

[0133] The exemplary embodiment of the present disclosure also provides a recommendation device based on federated learning. Figure 7 As shown, the recommendation device 700 based on federated learning includes the following program modules:

[0134] The receiving unit 701 is used to receive second data sets corresponding to each other sent by a plurality of local service devices, where the second data sets are obtained by the local service devices recalling data in the corresponding first data sets using the recall model;

[0135] A primary screening unit 702, configured to perform a primary screening on the obtained second data set, obtain a third data set corresponding to the local service device, and send the third data set to the corresponding local service device respectively;

[0136] The recommendation unit 703 is used to cluster and sort the obtained fourth data set according to the associated information of the user to be recommended, so as to obtain target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined ranking model to sort the data in the corresponding third data set.

[0137] In one implementation, the recommendation unit 703 is used to:

[0138] Determine the target comparison object according to the associated information of the user to be recommended;

[0139] Determine the correlation value between the data to be compared in each of the fourth data sets and the target comparison object, and determine the recommended candidate set to which each of the data to be compared belongs according to the correlation value corresponding to each of the data to be compared;

[0140] According to the ranking information of each recommendation candidate set, the target information recommended to the to-be-recommended user is determined.

[0141] In one implementation, the recommendation unit 703 is used to:

[0142] Determining, based on the association information of the user to be recommended, keywords of interest to the user to be recommended for the objects in the first data set;

[0143] A keyword portrait will be generated based on the interest keywords as the target comparison object.

[0144] In one implementation, the recommendation unit 703 is used to:

[0145] Determine the relevance ranges corresponding to the multiple recommendation candidate sets;

[0146] When it is determined that the correlation value corresponding to the first data to be compared is within the correlation range corresponding to the first recommendation candidate set, it is determined that the first data to be compared belongs to the first recommendation candidate set; the first data to be compared belongs to any of the fourth data sets.

[0147] In one embodiment, the device further comprises a display unit for:

[0148] Determining a target display format for the data in the target information according to the historical display information in the associated information of the user to be recommended;

[0149] The target information and the target display form are sent to a terminal device corresponding to the to-be-recommended user, so that the terminal device displays the data in the target information to the to-be-recommended user in the target display form.

[0150] The specific details of each part of the above-mentioned device have been described in detail in the implementation method of the method part. The undisclosed details can be found in the implementation method of the method part, so they will not be repeated here.

[0151] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the exemplary embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0152] The exemplary embodiment of the present disclosure also provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the above Figure 2-4 The recommendation method based on federated learning is shown.

[0153] In one embodiment, the computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing a computer program. The readable storage medium may be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk (HDD), solid-state drive (SSD), and the like. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing a computer program, such as a read-only memory, a NAND flash memory, and the like.

[0154] In one embodiment, the computer program product may be an intangible product including a computer program. Exemplarily, the computer program product may be implemented as a virtual digital product, such as a digital file storing an executable file, an installation package, etc. of the computer program.

[0155] The code of the computer program can be written in one or more programming languages. Programming languages ​​such as C language, Java, C++, etc. The program code can be executed entirely on the user computing device, or partially on the user computing device, or as a separate software package, or partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (e.g., an Internet connection provided by an operator).

[0156] The computer program may be carried or transmitted via electrical, magnetic, optical, electromagnetic, infrared, or other signals. The electronic device may convert the signal carrying the computer program into a digital signal, thereby running the computer program. When the computer program is run on an electronic device, its code is used to enable the electronic device to execute (more specifically, to enable the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure, such as the above-mentioned Figure 2-Figure 4 The recommendation method based on federated learning is shown in Figure 2 The recommendation method based on federated learning shown includes the following steps: Step 201: The local service device uses the recall model to recall the data in the corresponding first data set, obtains the corresponding second data set, and sends the second data set to the central service device. Step 202: The central service device receives the corresponding second data set sent by the local service device. Step 203: The central service device preliminarily screens the obtained second data set, obtains the corresponding third data set of the local service device, and sends the third data set to the corresponding local service device; Step 204: The local service device receives the corresponding third data set sent by the central service device; Step 205: The local service device uses the refined sorting model to sort the data in the corresponding third data set, obtains the corresponding fourth data set, and sends the fourth data set to the central service device; Step 206: The central service device receives the corresponding fourth data set sent by the local service device. Step 207: The central service device performs cluster sorting processing on the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended.

[0157] The above method steps are implemented by computer programs. Since the distributed recommendation system includes a central service device and multiple local service devices, and the recall model and the refined ranking model are deployed in the local service device. Therefore, the central service device can receive the second data sets corresponding to each other sent by multiple local service devices. The second data set is obtained by the local service device using the recall model to recall the data in the corresponding first data set; that is, in the embodiment of the present disclosure, federated learning is combined with the recall module in the recommendation system to achieve data screening in the recall stage without data going out of the domain. And, in order to reduce the amount of data processing, the central service device can perform a preliminary screening on the obtained second data set, obtain the third data set corresponding to the local service device, and then send the third data set to the corresponding local service device respectively. In other words, the refined ranking model is deployed in the local service device for sorting processing, which speeds up the recommendation sorting processing of the third data set, shortens the processing time of the entire recommendation system, and improves efficiency. Furthermore, in order to improve the accuracy of recommendation, the central service device also performs cluster sorting processing on the fourth data set obtained according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined sorting model to sort the data in the corresponding third data set. It can be seen that the embodiment of the present disclosure adds auxiliary sorting information on the basis of the sorting of the refined sorting module, thereby improving the recommendation accuracy of the recommendation system.

[0158] The exemplary embodiment of the present disclosure also provides an electronic device, which may be Figure 1 Local service equipment in Figure 1 The central service device in the electronic device may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the method steps of various exemplary embodiments of the present disclosure by executing the executable instructions.

[0159] Reference below Figure 8 , the electronic device is exemplarily described in the form of a general-purpose computing device. It should be understood that Figure 8 The electronic device 800 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0160] like Figure 8 As shown, the electronic device 800 may include: a processor 810 , a memory 820 , a bus 830 , an I / O (input / output) interface 840 , a network adapter 850 , and a display 880 .

[0161] The memory 820 may include a volatile memory, such as a RAM 821, a cache unit 822, and may also include a non-volatile memory, such as a ROM 823. The memory 820 may also include one or more program modules 824, such program modules 824 include but are not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or a combination thereof may include the implementation of a network environment. For example, the program module 824 may include each module in the above-mentioned device.

[0162] The processor 810 may include one or more processing units. For example, the processor 810 may include an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor and / or an NPU (Neural-Network Processing Unit), etc.

[0163] The processor 810 can be used to execute the executable instructions stored in the memory 820, such as the above-mentioned Figure 2-Figure 4 The recommendation method based on federated learning is shown in Figure 2 The proposed federated learning-based recommendation method includes the following steps:

[0164] Step 201: The local service device uses the recall model to recall the data in the corresponding first data set, obtains the corresponding second data set, and sends the second data set to the central service device. Step 202: The central service device receives the corresponding second data set sent by the local service device. Step 203: The central service device preliminarily screens the obtained second data set, obtains the corresponding third data set of the local service device, and sends the third data set to the corresponding local service device; Step 204: The local service device receives the corresponding third data set sent by the central service device; Step 205: The local service device uses the refined sorting model to sort the data in the corresponding third data set, obtains the corresponding fourth data set, and sends the fourth data set to the central service device; Step 206: The central service device receives the corresponding fourth data set sent by the local service device. Step 207: The central service device performs cluster sorting on the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended.

[0165] The above method steps are executed by processor 810. Since the distributed recommendation system includes a central service device and multiple local service devices, and the recall model and the refined ranking model are deployed in the local service device. Therefore, the central service device can receive the corresponding second data sets sent by multiple local service devices. The second data set is obtained by the local service device using the recall model to recall the data in the corresponding first data set; that is, in the embodiment of the present disclosure, federated learning is combined with the recall module in the recommendation system to achieve data screening in the recall stage without data going out of the domain. And, in order to reduce the amount of data processing, the central service device can perform a preliminary screening on the obtained second data set, obtain the third data set corresponding to the local service device, and then send the third data set to the corresponding local service device respectively. In other words, the refined ranking model is deployed in the local service device for sorting processing, which speeds up the recommendation sorting processing of the third data set, shortens the processing time of the entire recommendation system, and improves efficiency. Furthermore, in order to improve the accuracy of recommendation, the central service device also performs cluster sorting processing on the fourth data set obtained according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined sorting model to sort the data in the corresponding third data set. It can be seen that the embodiment of the present disclosure adds auxiliary sorting information on the basis of the sorting of the refined sorting module, thereby improving the recommendation accuracy of the recommendation system.

[0166] The bus 830 is used to realize the connection between different components of the electronic device 800, and may include a data bus, an address bus, and a control bus.

[0167] The electronic device 800 can communicate with one or more external devices 900 (eg, a keyboard, a mouse, an external controller, etc.) through the I / O interface 840 .

[0168] The electronic device 800 can communicate with one or more networks through the network adapter 850. For example, the network adapter 850 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, near field communication, etc. The network adapter 850 can communicate with other modules of the electronic device 800 through the bus 830.

[0169] The electronic device 800 may display a graphical user interface, such as target information, through the display 880 .

[0170] although Figure 8 Not shown, other hardware and / or software modules may also be provided in the electronic device 800, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0171] As can be seen from the above, the technical solution of the present disclosure can be implemented as a method, an apparatus, a system, a computer program product, a storage medium, an electronic device, etc. Those skilled in the art can understand that various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software, such as being respectively referred to as a "circuit", "module" or "system".

[0172] It should be understood that the present disclosure is not limited to the specific method steps or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. Those skilled in the art will easily think of other embodiments based on the specific embodiments provided by the present disclosure. Therefore, the specific embodiments provided by the present disclosure are only exemplary, and the scope and spirit of the present disclosure are indicated by the claims, and any variations, uses or adaptive changes of the present disclosure should be covered, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the technical field that are not disclosed in the present disclosure.

Claims

1. A recommendation method based on federated learning, characterized in that: Applied to a distributed recommendation system, the distributed recommendation system includes a central service device and multiple local service devices, and the local service devices are deployed with a recall model and a refined ranking model; the method includes: The central service device receives second data sets corresponding to each other sent by a plurality of local service devices, where the second data sets are obtained by the local service devices recalling data in the corresponding first data sets using the recall model; The central service device preliminarily screens the obtained second data set to obtain a third data set corresponding to the local service device, and sends the third data set to the corresponding local service device respectively; The central service device clusters and sorts the fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended; the fourth data set is obtained by the local service device sorting the data in the corresponding third data set using the refined sorting model.

2. The method according to claim 1, characterized in that The central service device performs clustering and sorting processing on the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended, including: The central service device determines the target comparison object according to the associated information of the user to be recommended; The central service device determines the correlation value between the data to be compared in each of the fourth data sets and the target comparison object, and determines the recommended candidate set to which each of the data to be compared belongs according to the correlation value corresponding to each of the data to be compared; The central service device determines the target information to be recommended to the to-be-recommended user according to the ranking information of each recommendation candidate set.

3. The method according to claim 2, characterized in that The central service device determines the target comparison object according to the associated information of the user to be recommended, including: The central service device determines, based on the association information of the user to be recommended, the keywords of interest of the user to be recommended to the objects in the first data set; The central service device will generate a keyword portrait based on the interest keywords as the target comparison object.

4. The method according to claim 3, characterized in that The central service device determines the recommendation candidate set to which each of the to-be-compared data belongs according to the correlation value corresponding to each of the to-be-compared data, including: The central service device determines the relevance ranges corresponding to the plurality of recommendation candidate sets respectively; When the central service device determines that the relevance value corresponding to the first data to be compared is within the relevance range corresponding to the first recommendation candidate set, it is determined that the first data to be compared belongs to the first recommendation candidate set; the first data to be compared belongs to any of the fourth data sets.

5. A recommendation method based on federated learning, characterized in that: Applied to a distributed recommendation system, the distributed recommendation system includes a central service device and multiple local service devices, and the local service devices are deployed with a recall model and a refined ranking model; the method includes: The local service device uses the recall model to perform recall processing on the data in the corresponding first data set to obtain the corresponding second data set, and sends the second data set to the central service device; The local service device receives a corresponding third data set sent by the central service device; the third data set is obtained by the central service device performing a preliminary screening on the corresponding second data set; The local service device uses the refined sorting model to sort the data in the corresponding third data set to obtain the corresponding fourth data set, and sends the fourth data set to the central service device, so that the central service device clusters and sorts the obtained fourth data set according to the associated information of the user to be recommended, and obtains the target information recommended to the user to be recommended.

6. The method according to claim 5, characterized in that The local service device uses the recall model to perform recall processing on each data in the corresponding first data set to obtain the corresponding second data set, including: The local service device calculates the recall score of each data in the first data set by using the recall model, and obtains a processed data set including the recall score; The local service device rearranges the data in the processed data set in a key-value pair manner to obtain a second data set.

7. A recommendation device based on federated learning, characterized in that: Applied to a distributed recommendation system, the distributed recommendation system includes a central service device and multiple local service devices, and the local service devices are deployed with a recall model and a refined ranking model; the device includes: A receiving unit, configured to receive second data sets corresponding to each other and sent by a plurality of local service devices, wherein the second data sets are obtained by the local service devices recalling data in the corresponding first data sets using the recall model; a primary screening unit, configured to perform a primary screening on the obtained second data set, obtain a third data set corresponding to the local service device, and send the third data set to the corresponding local service device respectively; The recommendation unit is used to cluster and sort the obtained fourth data set according to the associated information of the user to be recommended, so as to obtain target information recommended to the user to be recommended; the fourth data set is obtained by the local service device using the refined ranking model to sort the data in the corresponding third data set.

8. A recommendation device based on federated learning, characterized in that: Applied to a distributed recommendation system, the distributed recommendation system includes a central service device and multiple local service devices, and the local service devices are deployed with a recall model and a refined ranking model; the device includes: A recall unit, configured to use the recall model to perform recall processing on the data in the corresponding first data set, obtain the corresponding second data set, and send the second data set to the central service device; A receiving unit, configured to receive a corresponding third data set sent by the central service device; the third data set is obtained by the central service device performing a preliminary screening on the corresponding second data set; The processing unit is used to use the refined ranking model to sort the data in the corresponding third data set to obtain the corresponding fourth data set, and send the fourth data set to the central service device, so that the central service device can cluster and sort the obtained fourth data set according to the associated information of the user to be recommended, and obtain the target information recommended to the user to be recommended.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 4 or 5 to 6 is implemented.

10. An electronic device, characterized in that: include: processor; A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 4 or 5 to 6 by executing the executable instructions.

Images