Communication method and device

Abstract

The invention discloses a communication method and device. The method comprises the following steps: a first network element sends a first message to a second network element; the first message comprises the analysis identifier and is used for requesting a model for providing a service corresponding to the analysis identifier. The first network element receives a second message from a second network element, wherein the second message comprises the first identifier, the information of the M network elements and the information of the first model; wherein the first identifier is associated with the analysis identifier, and the first identifier corresponds to at least two longitudinal federal learning models; the M network elements use the at least two longitudinal federated learning models to provide services, M is a positive integer, and the at least two longitudinal federated learning models comprise the first model. By adopting the method, the first network element requests the second network element for providing the model of the service corresponding to the analysis identifier, and when the second network element determines to use the longitudinal federated learning model, the second network element sends the second message to the first network element, so that the first network element can initiate to use the longitudinal federated learning model to execute the analysis task according to the content of the second message.

Description

Technical Field

[0001] This application relates to the field of communications, and particularly to a communication method and apparatus. Background Art

[0002] Federated learning among multiple Network Data Analytics Function (NWDAF) network elements is a machine learning technology in the core network. It trains a Machine Learning (ML) model among multiple distributed entities holding local datasets without exchanging / sharing the local datasets. Compared with traditional centralized machine learning, this machine learning technology does not require uploading all local datasets to a single server, thus can solve key problems such as data privacy, data security, and data access rights.

[0003] According to the different characteristics of the data sources of the participating parties, federated learning can be classified into three categories: horizontal federated learning, vertical federated learning, and federated transfer learning. Currently, the 5G Core Network (5GC) only supports horizontal federated learning and performing analysis tasks through a horizontal federated learning model. How to implement performing analysis tasks through a vertical federated learning model is an issue worthy of attention. Summary of the Invention

[0004] Embodiments of this application provide a communication method and apparatus for implementing performing analysis tasks through a vertical federated learning model.

[0005] In a first aspect, this application provides a communication method. The execution subject of this method can be a first network element or a chip inside the first network element. The method includes: the first network element sends a first message to a second network element; the first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier; the first network element receives a second message from the second network element, and the second message includes a first identifier, information of M network elements, and information of a first model; wherein, the first identifier is associated with the analysis identifier, the first identifier corresponds to at least two vertical federated learning models; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, the at least two vertical federated learning models include the first model, and each of the M network elements corresponds to the at least two vertical federated learning models.

[0006] By adopting the above method, the first network element requests a model for providing the service corresponding to the analysis identifier from the second network element. When the second network element determines to use a vertical federated learning model, the second network element sends at least one of the first identifier, information of M network elements, and information of the first model to the first network element. Then, the first network element can initiate performing analysis tasks using the vertical federated learning model based on the above information.

[0007] In a possible design, the first network element determines at least one network element according to the information of the M network elements; the first network element sends a third message to the at least one network element, the third message includes the first identifier, and the third message is used to request the longitudinal federated learning model corresponding to the first identifier to provide services.

[0008] With the above design, the first network element can send a third message to at least one of the M network elements, that is, initiate an analysis task using the longitudinal federated learning model to at least one network element.

[0009] In a possible design, the first network element obtains a first intermediate result according to the first model; the first network element receives at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; the first network element determines an output according to the first intermediate result and the at least one intermediate result; the first network element sends the output to the service-consuming network element.

[0010] With the above design, the producer of the first model can be the main participant in the longitudinal federated task corresponding to the first identifier. Therefore, the first network element can collect each intermediate result and determine the output.

[0011] In a possible design, the first network element obtains a first intermediate result according to the first model; the first network element sends the first intermediate result to a third network element; the third network element is one of the at least one network element; the first network element receives an output from the third network element; the first network element sends the output to the service-consuming network element.

[0012] With the above design, the producer of the first model can be a secondary participant in the longitudinal federated task corresponding to the first identifier, and the producer of the model used by the third network element is the main participant in the longitudinal federated task corresponding to the first identifier. Therefore, the first network element can send the first intermediate result to the third network element, and the third network element determines the output.

[0013] In a possible design, before the first network element sends the first message to the second network element, the first network element receives a fourth message from the service-consuming network element, the fourth message includes the analysis identifier and / or information of the first object, wherein the first object corresponds to the analysis identifier; the fourth message is used to request the service corresponding to the analysis identifier.

[0014] In a possible design, the third message further includes the information of the first object.

[0015] In a possible design, the third message further includes an identifier of the first object in the vertical federated task corresponding to the first identifier;

[0016] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is a mapping relationship between the first object and the object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0017] With the above design, the first network element can convert the identifier of the first object into the identifier of the first object in the vertical federated task corresponding to the first identifier.

[0018] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier. If the information of the first object includes the network entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the network entity identifier of the first object and the first mapping relationship. If the information of the first object includes the user entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the user entity identifier of the first object and the second mapping relationship. With the above design, the first network element can convert the network entity identifier or user entity identifier of the first object into the identifier of the first object in the vertical federated task corresponding to the first identifier, that is, the identifier of the first object in the vertical federated task.

[0019] In a possible design, the user entity identifier is SUPI.

[0020] In a possible design, the second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; or, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or, the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0021] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0022] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier; wherein, the j-th network element is a user of the j-th vertical federated learning model, and the at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0023] In a second aspect, the present application provides a communication method. The execution subject of this method can be a second network element or a chip inside the second network element. The method includes: the second network element receives a first message from a first network element; the first message includes an analysis identifier, and the first message is used to request a model for providing a service corresponding to the analysis identifier; the second network element sends a second message to the first network element, and the second message includes a first identifier, the information of M network elements, and the information of a first model; wherein, the first identifier is associated with the analysis identifier, and there are at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model, and each of the M network elements corresponds to the at least two vertical federated learning models.

[0024] By using the above method, the first network element requests a model for providing a service corresponding to the analysis identifier from the second network element. When the second network element determines to use a vertical federated learning model, the second network element sends at least one of the first identifier, the information of M network elements, and the information of the first model to the first network element. Furthermore, the first network element can initiate an analysis task using the vertical federated learning model based on the above information.

[0025] In a possible design, before the second network element sends the second message to the first network element, the second network element determines at least one of the first identifier, the information of the M network elements, and the information of the first model according to the analysis identifier.

[0026] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity identifier in the vertical federated task corresponding to the first identifier.

[0027] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is the mapping relationship between the first object and the object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0028] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0029] In a possible design, the second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; or, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or, the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0030] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0031] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier; where the j-th network element is the user of the j-th vertical federated learning model, and at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0032] In a third aspect, the present application provides a communication method. The execution subject of this method can be a first network element or a chip inside the first network element. The method includes: the first network element sends a first message to the second network element; the first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier; the first network element receives a second message from the second network element, the second message includes a second identifier and / or first indication information, the first indication information indicates that the model for providing the service corresponding to the analysis identifier is a vertical federated learning model, and the second identifier is used to identify the first message; the first network element sends a third message to the second network element, and the third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

[0033] Using the above method, the first network element requests a model for providing the service corresponding to the analysis identifier from the second network element. When the second network element determines to use a vertical federated learning model, the second network element sends the second identifier and / or the first indication information to the first network element. That is, the second network element does not provide the information of the model for the first network element, but uses the first indication information to notify the first network element that the model for providing the service corresponding to the analysis identifier is a vertical federated learning model, and uses the second identifier to enable the second network element to request the service corresponding to the analysis identifier, that is, initiate an analysis task using the vertical federated learning model.

[0034] In a possible design, the first network element receives an output from the second network element; the first network element sends the output to a service consumption network element.

[0035] In a possible design, before the first network element sends the first message to the second network element, the first network element receives a fourth message from the service consumption network element, the fourth message includes the analysis identifier and / or information of a first object, where the first object corresponds to the analysis identifier; the fourth message is used to request the service corresponding to the analysis identifier.

[0036] In a possible design, the third message further includes the information of the first object.

[0037] In a possible design, the second message further indicates or includes a first mapping relationship, and the first mapping relationship is a mapping relationship between the first object and the object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0038] With the above design, the first network element can convert the identifier of the first object into the identifier of the first object in the vertical federated task corresponding to the first identifier.

[0039] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier. If the information of the first object includes the network entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the network entity identifier of the first object and the first mapping relationship. If the information of the first object includes the user entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the user entity identifier of the first object and the second mapping relationship.

[0040] With the above design, the first network element can convert the network entity identifier or the user entity identifier of the first object into the identifier of the first object in the vertical federated task corresponding to the first identifier, that is, the identifier of the first object in the vertical federated task.

[0041] In a fourth aspect, the present application provides a communication method. The execution subject of this method can be a second network element or a chip inside the second network element. The method includes: The second network element receives a first message from the first network element. The first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier. The second network element sends a second message to the first network element. The second message includes a second identifier and / or a first indication information. The first indication information indicates that the model for providing the service corresponding to the analysis identifier is a vertical federated learning model. The second identifier is used to identify the first message. The second network element receives a third message from the first network element. The third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

[0042] With the above method, when the first network element requests a model for providing the service corresponding to the analysis identifier from the second network element, and the second network element determines to use the vertical federated learning model, the second network element sends the second identifier and / or the first indication information to the first network element. Furthermore, the first network element can determine to use the vertical federated learning model according to the first indication information, and send the third message according to the second identifier, that is, initiate an analysis task using the vertical federated learning model.

[0043] In a possible design, before the second network element sends a second message to the first network element, the second network element determines information of a first model according to the analysis identifier; the second network element stores a correspondence between the second identifier and the information of the first model.

[0044] Using the above method, the second network element stores a correspondence between the second identifier and the information of the first model, and thus can determine the information of the first model according to the second identifier in the third message, and provide a service corresponding to the analysis identifier based on the first model, that is, initiate an analysis task using a vertical federated learning model.

[0045] In a possible design, when the second network element determines the information of the first model according to the analysis identifier, the second network element determines the information of the first model, information of M network elements, and a first identifier according to the analysis identifier; when the second network element stores a correspondence between the second identifier and the information of the first model, the second network element stores a correspondence between the second identifier and the information of the first model, the information of the M network elements, and the first identifier; wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models correspond to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model, and each of the M network elements corresponds to the at least two vertical federated learning models.

[0046] In a possible design, after the second network element receives a third message from the first network element, the second network element determines the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; the second network element determines at least one network element according to the information of the M network elements; the second network element sends a fifth message to the at least one network element, the fifth message includes the first identifier, and the fifth message is used to request a vertical federated learning model corresponding to the first identifier to provide a service; the second network element obtains a first intermediate result according to the first model; the second network element receives at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; the second network element determines the output according to the first intermediate result and the at least one intermediate result.

[0047] Using the above design, the second network element is the main participant in the vertical federated task corresponding to the first identifier.

[0048] In a possible design, after the second network element receives a third message from the first network element, the second network element determines information of the first model, information of the M network elements, and the first identifier according to the second identifier; the second network element determines at least one network element according to the information of the M network elements; the second network element sends a fifth message to the at least one network element, where the fifth message includes the first identifier, and the third message is used to request a vertical federated learning model corresponding to the first identifier to provide services; the second network element obtains a first intermediate result according to the first model; the second network element sends the first intermediate result to the third network element; the second network element receives an output from the third network element.

[0049] With the above design, the second network element is the slave main participant of the vertical federated task corresponding to the first identifier. The producer of the model used by the third network element is the slave main participant of the vertical federated task corresponding to the first identifier.

[0050] In a possible design, the second network element sends the output to the first network element.

[0051] In a possible design, the third message further includes information of a first object and / or an identifier of the first object in the vertical federated task corresponding to the first identifier.

[0052] In a possible design, the second message further indicates a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is an identifier of a network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity in the vertical federated task corresponding to the first identifier.

[0053] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is a mapping relationship between a first object and an object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between an identifier of the first object and an identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0054] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identity of a network entity in the vertical federated task corresponding to the first identity, and the second mapping relationship is the identity of a user entity in the vertical federated task corresponding to the first identity. It can be understood that the first mapping relationship is the relationship between the network entity identity and the identity in the vertical federated task corresponding to the first identity, and the second mapping relationship is the relationship between the user entity identity and the identity in the vertical federated task corresponding to the first identity.

[0055] In a possible design, the information of the M network elements includes at least one of the network entity identities of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0056] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identity. Wherein, the j-th network element is the user of the j-th vertical federated learning model, and at least two vertical federated learning models corresponding to the first identity include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0057] In a fifth aspect, the present application provides a communication device, and the method includes: a transceiver unit and a processing unit; the transceiver unit is used for sending and receiving messages; the processing unit is used for sending a first message to a second network element through the transceiver unit; and receiving a second message from the second network element. The first message includes an analysis identity, and the first message is used to request a model that provides services corresponding to the analysis identity. The second message includes a first identity, information of M network elements, and information of a first model. Wherein, the first identity is associated with the analysis identity, and at least two vertical federated learning models corresponding to the first identity; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model, and each of the M network elements corresponds to the at least two vertical federated learning models.

[0058] In a possible design, the processing unit is used to determine at least one network element according to the information of the M network elements; the transceiver unit is used to send a third message to the at least one network element, and the third message includes the first identity, and the third message is used to request the vertical federated learning model corresponding to the first identity to provide services.

[0059] In a possible design, the processing unit is configured to obtain a first intermediate result according to the first model; the transceiver unit is configured to receive at least one intermediate result from at least one network element, where the at least one network element corresponds to the at least one intermediate result one by one; the processing unit is configured to determine an output according to the first intermediate result and the at least one intermediate result; and the transceiver unit is configured to send the output to a service consumption network element.

[0060] In a possible design, the processing unit is configured to obtain a first intermediate result according to the first model; the transceiver unit is configured to send the first intermediate result to a third network element, receive an output from the third network element, and send the output to a service consumption network element; and the third network element is one of the at least one network element.

[0061] In a possible design, the transceiver unit is configured to receive a fourth message from the service consumption network element before sending the first message to the second network element, where the fourth message includes the analysis identifier and information about a first object, and the first object corresponds to the analysis identifier; and the fourth message is used to request a service corresponding to the analysis identifier.

[0062] In a possible design, the third message further includes information about the first object.

[0063] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is a mapping relationship between the first object and the object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0064] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier; it can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0065] In a possible design, the third message further includes the identifier of the first object in the vertical federated task corresponding to the first identifier; the second message further indicates or includes a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is the identifier of the network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity identifier in the vertical federated task corresponding to the first identifier; the processing unit is configured to, if the information of the first object includes the network entity identifier of the first object, determine the identifier of the first object in the vertical federated task corresponding to the first identifier according to the network entity identifier of the first object and the first mapping relationship; if the information of the first object includes the user entity identifier of the first object, determine the identifier of the first object in the vertical federated task corresponding to the first identifier according to the user entity identifier of the first object and the second mapping relationship.

[0066] In a possible design, the user entity identifier is SUPI.

[0067] In a possible design, the second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; or, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or, the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0068] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0069] In a possible design, the jth network element is any one of the M network elements, and the role information of the jth network element indicates the role of the producer of the jth vertical federated learning model in the vertical federated task corresponding to the first identifier; where the jth network element is the user of the jth vertical federated learning model, and at least two vertical federated learning models corresponding to the first identifier include the jth vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0070] Sixth aspect, the present application provides a communication device, which includes a transceiver unit and a processing unit; the transceiver unit is used to send and receive messages, and the processing unit is used to receive a first message from a first network element through the transceiver unit and send a second message to the first network element; the first message includes an analysis identifier, and the first message is used to request a model for providing a service corresponding to the analysis identifier; the second message includes a first identifier, information of M network elements, and information of a first model; wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model, and each of the M network elements corresponds to the at least two vertical federated learning models.

[0071] In a possible design, the processing unit is configured to determine at least one of the first identifier, the information of the M network elements, and the information of the first model according to the analysis identifier before sending the second message to the first network element.

[0072] In a possible design, the second message further indicates or includes a first mapping relationship, and the first mapping relationship is a mapping relationship between a first object and an object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0073] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is the identifier of a network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of a user entity in the vertical federated task corresponding to the first identifier; it can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0074] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is the identifier of a network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of a user entity in the vertical federated task corresponding to the first identifier.

[0075] In a possible design, the second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; or, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or, the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0076] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0077] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier; where the j-th network element is a user of the j-th vertical federated learning model, and at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0078] In a seventh aspect, the present application provides a communication device, and the method includes: a transceiver unit and a processing unit; the transceiver unit is configured to send and receive messages, and the processing unit is configured to send a first message to a second network element through the transceiver unit, receive a second message from the second network element, and the second network element sends a third message; the first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier; the second message includes a second identifier and / or first indication information, the first indication information indicates that the model for providing the service corresponding to the analysis identifier is a vertical federated learning model, and the second identifier is used to identify the first message; the third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

[0079] In a possible design, the transceiver unit is configured to receive an output from the second network element; and send the output to a service consumption network element.

[0080] In a possible design, the transceiver unit is configured to receive a fourth message from the service consumption network element before sending the first message to the second network element, where the fourth message includes the analysis identifier and information of a first object, and where the first object corresponds to the analysis identifier; the fourth message is used to request the service corresponding to the analysis identifier.

[0081] In a possible design, the third message further includes the information of the first object.

[0082] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is a mapping relationship between a first object and an object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0083] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0084] In a possible design, the third message further includes the identifier of the first object in the vertical federated learning model. The second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated learning model, and the second mapping relationship is the identifier of the user entity in the vertical federated learning model. The processing unit is configured to, if the information of the first object includes the network entity identifier of the first object, determine the identifier of the first object in the vertical federated learning model according to the network entity identifier of the first object and the first mapping relationship; if the information of the first object includes the user entity identifier of the first object, determine the identifier of the first object in the vertical federated learning model according to the user entity identifier of the first object and the second mapping relationship.

[0085] In an eighth aspect, the present application provides a communication device, and the method includes: a transceiver unit and a processing unit. The transceiver unit is configured to send and receive messages. The processing unit is configured to receive, through the transceiver unit, a first message from a first network element, send a second message to the first network element, and receive a third message from the first network element. The first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier. The second message includes a second identifier and / or a first indication information, where the first indication information indicates that the model for providing the service corresponding to the analysis identifier is a vertical federated learning model, and the second identifier is used to identify the first message. The third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

[0086] In a possible design, the processing unit is configured to, before sending a second message to the first network element, the second network element determines information of a first model according to the analysis identifier; and save the correspondence between the second identifier and the information of the first model.

[0087] In a possible design, the processing unit is configured to, when determining the information of the first model according to the analysis identifier, the second network element determines the information of the first model, the information of M network elements, and a first identifier according to the analysis identifier; when the second network element saves the correspondence between the second identifier and the information of the first model, save the correspondence between the second identifier and the information of the first model, the information of the M network elements, and the first identifier; wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models correspond to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model, and each of the M network elements corresponds to the at least two vertical federated learning models.

[0088] In a possible design, the processing unit is configured to, after receiving a third message from the first network element, determine the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; determine at least one network element according to the information of the M network elements; the transceiver unit is configured to send a fifth message to the at least one network element, the fifth message includes the first identifier, and the fifth message is used to request a vertical federated learning model corresponding to the first identifier to provide services; the processing unit is configured to obtain a first intermediate result according to the first model; the transceiver unit is configured to receive at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; the processing unit is configured to determine the output according to the first intermediate result and the at least one intermediate result.

[0089] In a possible design, the processing unit is configured to, after receiving a third message from the first network element, determine the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; determine at least one network element according to the information of the M network elements; the transceiver unit is configured to send a fifth message to the at least one network element, the fifth message includes the first identifier, and the third message is used to request a vertical federated learning model corresponding to the first identifier to provide services; the processing unit is configured to obtain a first intermediate result according to the first model; the transceiver unit is configured to send the first intermediate result to the third network element; the second network element receives the output from the third network element.

[0090] In a possible design, the transceiver unit is configured to send the output to the first network element.

[0091] In a possible design, the third message further includes information of a first object, and / or an identifier of the first object in the vertical federated task corresponding to the first identifier.

[0092] In a possible design, the second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is an identifier of a network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity in the vertical federated task corresponding to the first identifier.

[0093] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is a mapping relationship between a first object and an object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0094] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is an identifier of a network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0095] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0096] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier. Wherein, the j-th network element is a user of the j-th vertical federated learning model, and at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0097] In a ninth aspect, the present application provides a communication device, which may be a first device, or a module or unit (e.g., a chip, or a chip system, or a circuit) corresponding to each of the methods / operations / steps / actions described in any one of the first to fourth aspects executed in the first device, or is capable of being used in matching with the first device.

[0098] In a tenth aspect, the present application provides a communication device, including at least one processing element and at least one storage element, where the at least one storage element is used to store programs and data, and the at least one processing element is used to read and execute the programs and data stored in the storage element, so that the method described in any one of the first to fourth aspects of the present application is implemented.

[0099] In an eleventh aspect, the present application further provides a computer program, which, when running on a computer, causes the computer to execute the method described in any one of the first to fourth aspects of the present application.

[0100] In a twelfth aspect, the present application provides a communication device, which includes: an interface circuit and at least one processor; the interface circuit is used to provide input and / or output of programs or instructions for the at least one processor; the at least one processor is used to execute the programs or instructions so that the communication device can implement the method described in any one of the first to fourth aspects of the present application.

[0101] In a possible manner, the communication device includes the at least one memory, and the at least one memory is used to store the programs or instructions.

[0102] In a thirteenth aspect, the present application provides a computer storage medium, in which a software program is stored, and when the software program is read and executed by one or more processors, the method described in any one of the first to fourth aspects of the present application can be implemented.

[0103] In a fourteenth aspect, the present application provides a computer program product containing instructions, which, when running on a computer, causes the computer to execute the method described in any one of the first to fourth aspects of the present application.

[0104] In a fifteenth aspect, the present application provides a chip system, which includes at least one chip and a memory, and the at least one chip is used to read and execute the programs stored in the memory to implement the method described in any one of the first to fourth aspects of the present application.

[0105] In a sixteenth aspect, the present application provides a communication system, which includes a first network element and a second network element. The first network element executes the method described in any one of the above first aspect and third aspect, and the second network element executes the method described in any one of the above second aspect and fourth aspect. Description of the Drawings

[0106] Figure 1 It is a schematic diagram of a 5G network architecture based on a service-based architecture in the present application;

[0107] Figure 2 It is a schematic diagram of the process of vertical federated learning training in the present application;

[0108] Figure 3 It is an overview flowchart of a communication method in the present application;

[0109] Figure 4 It is one of the flowcharts for obtaining an output in the present application;

[0110] Figure 5 It is another flowchart for obtaining an output in the present application;

[0111] Figure 6 It is a third flowchart for obtaining an output in the present application;

[0112] Figure 7 It is an overview flowchart of another communication method in the present application;

[0113] Figure 8 It is a flowchart of the first network element and at least one network element performing an analysis task through a vertical federated learning model in the present application;

[0114] Figure 9 It is a flowchart of the second network element and at least one network element performing an analysis task through a vertical federated learning model in the present application;

[0115] Figure 10 It is a flowchart of yet another communication method in the present application;

[0116] Figure 11 It is a flowchart of still another communication method in the present application;

[0117] Figure 12 It is a schematic diagram of the structure of a communication device in the present application;

[0118] Figure 13 It is a schematic diagram of the structure of another communication device in the present application. Detailed Embodiments

[0119] The specific implementation manners of the present application will be described by way of example with reference to the accompanying drawings in the embodiments of the present application. However, the implementation manners of the present application may also include combining these embodiments without departing from the spirit or scope of the present application, such as adopting other embodiments and making structural changes. Therefore, the detailed description of the following embodiments should not be understood in a limiting sense. The terms used in the embodiments of the present application are only used to explain the specific embodiments of the present application, rather than intended to limit the present application.

[0120] The embodiments of the present application can be applied to various communication systems, such as: Global System for Mobile Communications (GSM) system, Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, General Packet Radio Service (GPRS), Long Term Evolution (LTE) system, LTE Frequency Division Duplex (FDD) system, LTE Time Division Duplex (TDD), Universal Mobile Telecommunications System (UMTS), Worldwide Interoperability for Microwave Access (WIMAX) communication system, 5th generation (5G) system or New Radio (NR), or applied to future communication systems or other similar communication systems, etc.

[0121] Figure 1 It is a schematic diagram of a 5G network architecture based on a service-based architecture. Figure 1In the 5G network architecture shown, it may include terminal devices, access network devices, and core network devices. The terminal devices access the data network (DN) through the access network devices and the core network devices. Among them, the core network devices include multiple network functions (NFs) or network elements. For example, it includes some or all of the following network elements: unified data management (UDM) network element, unified data repository (UDR) network element, application function (AF) network element, policy control function (PCF) network element, access and mobility management function (AMF) network element, session management function (SMF) network element, user plane function (UPF) network element, network data analytics function (NWDAF) network element, network repository function (NRF) network element (not shown in the figure), etc.

[0122] The access network device may be a radio access network (RAN) device. For example: base station, evolved NodeB (eNodeB), transmission reception point (TRP), next generation NodeB (gNB) in the 5G mobile communication system, next generation NodeB in the 6th generation (6G) mobile communication system, base station in the future mobile communication system, or access node in the wireless fidelity (WiFi) system, etc.; it may also be a module or unit that completes some functions of the base station. For example, it may be a central unit (CU) or a distributed unit (DU). The radio access network device may be a macro base station, a micro base station, or an indoor station, and may also be a relay node or a donor node, etc. The embodiments of the present application do not limit the specific technologies and specific device forms adopted by the radio access network device.

[0123] The terminal device can be a user equipment (UE), a mobile station, a mobile terminal, etc. The terminal device can be widely applied to various scenarios, such as device-to-device (D2D), vehicle to everything (V2X) communication, machine-type communication (MTC), internet of things (IOT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearables, smart transportation, smart city, etc. The terminal device can be a mobile phone, a tablet computer, a computer with wireless transceiver function, a wearable device, a vehicle, an urban air vehicle (such as an unmanned aircraft, a helicopter, etc.), a ship, a robot, a robotic arm, a smart home device, etc.

[0124] The access network device and the terminal device can be in a fixed position or movable. The access network device and the terminal device can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on the water surface; they can also be deployed on airplanes, balloons, and artificial satellites in the air. The embodiments of the present application do not limit the application scenarios of the access network device and the terminal device.

[0125] The following briefly introduces some core network devices:

[0126] The AMF network element, abbreviated as AMF, includes functions such as performing mobility management, access authentication / authorization, etc. In addition, it is also responsible for transmitting user policies between the terminal device and the PCF.

[0127] The SMF network element, abbreviated as SMF, includes functions such as performing session management, executing the control policies issued by the PCF, selecting the UPF, and allocating the internet protocol (IP) address of the terminal device.

[0128] The UPF network element, abbreviated as UPF, as an interface to the data network, includes functions such as completing user plane data forwarding, session / flow-level billing statistics, and bandwidth limitation.

[0129] The UDM network element, abbreviated as UDM, includes functions such as performing management of subscription data and user access authorization.

[0130] The UDR network element, abbreviated as UDR, includes functions such as accessing and storing types of data such as subscription data, policy data, and application data.

[0131] The NEF network element, abbreviated as NEF, is used to support the opening of capabilities and events.

[0132] The AF network element, abbreviated as AF, transmits the requirements from the application side to the network side. For example, it can be quality of service (QoS) requirements or user status event subscriptions. The AF can be a third-party functional entity or an application server deployed by the operator.

[0133] The PCF network element, abbreviated as PCF, includes policy control functions such as session and traffic flow level charging, QoS bandwidth guarantee, mobility management, and terminal device policy decision-making.

[0134] The NRF network element, abbreviated as NRF, can be used to provide network element discovery functions. Based on the requests of other network elements, it provides network element information corresponding to the network element type. The NRF network element also provides network element management services, such as network element registration, update, deregistration, and network element status subscription and push.

[0135] The NWDAF network element, abbreviated as NWDAF, has the function of collecting data (including one or more of terminal device data, access network device data, core network element data, and third-party application device data). Among them, these data can be the data of the terminal device, access network device, core network element, or third-party application device itself, or the data of the terminal device on the access network device, the core network element, or the third-party application device. The NWDAF also has functions such as model training, data analysis, and model inference. For example, it performs data analysis based on the collected data and outputs the data analysis results for use in network, network management device, and application execution policy decision-making. In the embodiments of this application, one NWDAF can be a separate network element or co-located with other network elements. For example, the NWDAF can be set in the PCF network element or the AMF network element.

[0136] In the 3rd generation partnership project (3GPP) Release 17, the training function and inference function of the NWDAF are split. One NWDAF can only support the model training function, or only support the data inference function, or support both the model training function and the data inference function.

[0137] Among them, the NWDAF that supports the model training function can be used to train a machine learning (ML) model or an artificial intelligence (AI) model and expose new training services; for example, providing a trained AI model or ML model. The NWDAF that supports the model training function can also be referred to as the training NWDAF, or the NWDAF that supports the model training logical function (MTLF), simply referred to as MTLF. Exemplarily, the MTLF can perform model training based on the acquired data to obtain a trained model.

[0138] The NWDAF that supports the data inference function can be used to infer and export analysis information and expose analysis services. The NWDAF that supports the data inference function can also be referred to as the inference NWDAF, or the NWDAF that supports the analytics logical function (AnLF), simply referred to as AnLF. Exemplarily, the AnLF can request a model through the model subscription (MLModelProvision_Subscribe) service or a message from the MTLF, and this model can be obtained by the MTLF training based on the relevant data of the model. Furthermore, the AnLF can input the input data into the trained model to obtain analysis results or inference data.

[0139] It can be understood that the MTLF can be understood as the NWDAF that at least supports the model training function. As a possible implementation method, the MTLF can also support the data inference function. The AnLF can be understood as the NWDAF that at least supports the data inference function. As a possible implementation method, the AnLF can also support the model training function.

[0140] It can be understood that the above network elements are examples of one implementation method, and this application does not exclude that in a 6G or newer wireless communication system, there may be network elements or devices with the functions of the above network elements having other names, or having other forms.

[0141] It can be understood that the above network elements or functions can be either network elements in hardware devices, or software functions running on dedicated hardware, or virtualized functions instantiated on a platform (such as a cloud platform). As a possible implementation method, the above network elements or functions can be implemented by one device, or jointly implemented by multiple devices, or can also be a functional module within one device, and the embodiments of this application do not make specific limitations on this.

[0142] Figure 1Nudr, Npcf, Namf, Nudm, Nsmf, Naf, and Nnwdaf are service - oriented interfaces provided by the above - mentioned UDR, PCF, AMF, UDM, SMF, AF, and NWDAF respectively, which are used to call corresponding service - oriented operations.

[0143] The following explains the basic technical concepts involved in this application:

[0144] 1. Federated Learning

[0145] Federated learning (FL) is a machine - learning framework that can effectively help multiple users conduct data usage and machine - learning modeling while meeting the requirements of user privacy protection, data security, and government regulations. As a distributed machine - learning paradigm, federated learning can effectively solve the data - silo problem, conduct joint modeling without sharing user data, and thus technically break data silos and achieve artificial - intelligence (AI) collaboration.

[0146] 2. Vertical Federated Learning (VFL)

[0147] To train data from different domains, distributed model training can be achieved by means of vertical federated learning (VFL). As a machine - learning technology, VFL can be used to solve model training and inference in the case where each member is reluctant to share raw data. It is applicable to the situation where there is a large overlap in the training - sample identification (ID) of participants, while there is a small overlap in the data features of participants. VFL combines different data features of the common samples of multiple participants for federated learning. That is, the training data of each participant is vertically divided, so it is called vertical federated learning.

[0148] Taking the linear - regression algorithm as an example, the training process of vertical federated learning is as Figure 2 shown.

[0149] Client A has a dataset Client B has a dataset where y i is the label data, and the model to be trained is as follows:

[0150]

[0151] Assume the objective function for linear regression is as follows,

[0152]

[0153] Among them, L is the loss function, specifically as follows:

[0154]

[0155] Due to the original data D on client A A and the D on client B B cannot be aggregated together, so it is impossible to train based on the traditional centralized training method, but it can be trained based on the vertical federated training method, as follows:

[0156] Let Then the transformation of L is as follows:

[0157]

[0158] Let Then

[0159] L = L A + L B + L AB (Formula 5)

[0160] Let the residual Then the gradient of L with respect to Θ A and Θ B is as follows:

[0161]

[0162]

[0163] Correspondingly, the model parameters are updated as follows:

[0164]

[0165]

[0166] Then the training process of vertical federated learning is as follows:

[0167] Step 1, client A and client B respectively initialize the model parameters Θ A and Θ B ;

[0168] Step 2, client A calculates A and L based on Θ A , and then sends them to client B;

[0169] Step 3, client B calculates B based on Θ Further based on and y i calculate d i 、L AB 、L B ,and finally based on L A 、L AB 、L B calculate to obtain L. Client B sends d i to client A;

[0170] Step 4, client A and client B each calculate based on d i respectively and Then, respectively based on and update the model parameters Θ A and Θ B .

[0171] Among them, steps 2 to 4 are executed in a loop until the model training end condition is reached, such as the number of iterations reaches a set threshold (such as 10,000 times) or the value of the loss function L is less than a set threshold (such as 0.001).

[0172] Through the above technology, the interaction of raw data between different domains is avoided, and a business experience model can also be trained. After the training is completed, for a set of data In the inference stage, client A and client B respectively calculate the local inference results (also called intermediate results) based on the trained model parameters Θ A and Θ B Then client A sends the local inference result and to client B, and client B finally determines the inference result (also called the prediction result, or analysis output, or output)

[0173] For the sake of convenience of description, Θ A is used below to represent the vertically federated learning model trained by client A, and Θ B is used to represent the vertically federated learning model trained by client B.

[0174] 3. Participants in Vertically Federated Learning

[0175] In this application, the participants in vertically federated learning can also be called the participating devices in vertically federated learning, or the participants in vertical federation tasks, or the participants in vertically federated learning tasks. This application does not make any restrictions on this. This application takes the participants in vertical federation tasks as an example for illustration.​

[0176] In vertical federated learning, there are multiple types of participants, such as UEs, NWDAFs, AFs, etc. For example, the participants in a vertical federated task include the aforementioned client A and client B. Optionally, the participants in a vertical federated task can also include more clients, without specific limitation. For another example, the aforementioned client A and client B can be different AFs or different NWDAFs respectively, or be an AF and an NWDAF respectively.

[0177] Exemplarily, the main roles of the participants in a vertical federated task are specifically as follows:

[0178] Primary participant: An entity that is responsible for providing the labeled data of the training data for a vertical federated task during the vertical federated learning model training phase. Among them, the primary participant can determine the final output based on multiple intermediate results. For example, the aforementioned client B is the primary participant. The primary participant can also provide part of the training data for the vertical federated task.

[0179] Secondary participant: An entity that is responsible for providing part of the training data for a vertical federated task. Among them, the secondary participant can determine intermediate results through the corresponding vertical federated learning model, but cannot determine the final output. For example, the aforementioned client A is the secondary participant.

[0180] Optionally, the participants in a vertical federated task can also include a coordinator, where the coordinator is responsible for coordinating the participants in training a vertical federated learning model.

[0181] 4. Analysis ID

[0182] The analysis ID can be used to indicate an analysis service, which is abbreviated as a service. This service is related to the model, that is, the model can be used to execute this service. Or it can be described as that the analysis ID is related to the model, that is, the model is used to execute the service corresponding to the analysis ID.

[0183] Or it can also be understood that MTLF is related to the analysis ID, that is, the model provided by MTLF supports the execution of the service corresponding to this analysis ID. Exemplarily, MTLF can be related to one or more analysis IDs. It can be understood that this MTLF can provide models for the services corresponding to each of the one or more analysis IDs. For example, MTLF1 is related to analysis ID 1 and analysis ID 2, that is, MTLF1 corresponds to analysis ID 1 and analysis ID 2, then MTLF1 can provide models for the services corresponding to analysis ID 1 and analysis ID 2.

[0184] In this application, a vertical federated task refers to a task corresponding to vertical federated learning. A vertical federated task can also be referred to as a vertical federated activity or a vertical federated learning task. A vertical federated task can be associated with an analysis identifier, that is, the vertical federated learning model provided by the vertical federated task can provide services corresponding to the analysis identifier. Or, it can also be understood that the identifier of the vertical federated task has an association relationship with the analysis identifier.

[0185] A vertical federated task can include training and / or inference. Among them, training can also be referred to as model training or vertical federated learning model training, which is used to obtain a vertical federated learning model. Inference refers to using the vertical federated learning model obtained in the training stage to perform inference or execute an inference task (or a prediction task or an analysis task). Inference can also be referred to as model inference, or vertical federated learning model inference, or model prediction, or vertical federated learning model prediction. In this application, inference and prediction can be used interchangeably.

[0186] The identifier of a vertical federated task is used to identify a vertical federated task, and to determine the vertical federated learning model corresponding to the vertical federated task, and / or the participant information of the vertical federated task, and / or the user information of the vertical federated learning model.

[0187] It can be understood that there can be multiple vertical federated learning models corresponding to a vertical federated task. Furthermore, there can be multiple participants in the vertical federated task, that is, each vertical federated learning model corresponds to a corresponding participant. It can be understood that the participants in the vertical federated task refer to some or all of the participants in the vertical federated task. The participants in the vertical federation jointly participate in the vertical federated task and obtain independent vertical federated learning models respectively. The participants in the vertical federation can include a coordinator or may not include a coordinator.

[0188] Similarly, there can be multiple users of the vertical federated learning model, that is, each vertical federated learning model corresponds to a corresponding user. It can be understood that the user of the vertical federated learning model refers to an entity that uses the vertical federated learning model obtained in the training stage to perform inference or execute an inference task (or a prediction task or an analysis task). Inference can also be referred to as model inference, or vertical federated learning model inference, or model prediction, or vertical federated learning model prediction.

[0189] In a possible implementation, the participants in the vertical federated task can be the same as the users of the vertical federated learning model. In this case, the two descriptions of the participants in the vertical federated task and the users of the vertical federated learning model can be replaced with each other. For example, the aforementioned client A and client B can both be the participants in the vertical federated task and the users of the vertical federated learning model. That is, client A participates in the training of the vertical federated learning model Θ A, a vertical federated learning model Θ can also be used A . Similarly, client B not only participates in the training of the vertical federated learning model Θ B , but can also use the vertical federated learning model Θ B .

[0190] In another possible implementation, the participants in the vertical federated task and the users of the vertical federated learning model can also be different. For example, the aforementioned client A and / or client B only act as participants in the vertical federated task and not as users of the vertical federated learning model. For example, client A only participates in the training of the vertical federated learning model Θ A , and the user of the vertical federated learning model Θ A is not client A, or client A does not have the permission to use the vertical federated learning model Θ A . For example, the user of the vertical federated learning model Θ A is client C, that is, client C can use the vertical federated learning model Θ A , and optionally, client A cannot use the vertical federated learning model Θ A . Among them, if the role of client A in the vertical federated task is a secondary participant, then the role of client C is also a secondary participant. That is, if client B not only participates in the training of the vertical federated learning model Θ B , but can also use the vertical federated learning model Θ B , and the role of client B in the vertical federated task is the primary participant, then the intermediate results calculated by client C need to be sent to client B.

[0191] Exemplarily, the participant information in the vertical federated task may include but is not limited to at least one of the network entity identifier of the participant, the subscription associated address of the participant, and the role information of the participant. The user information of the vertical federated learning model may include but is not limited to at least one of the network entity identifier of the user, the subscription associated address of the user, and the role information of the user. Among them, the role information of the user is the same as the role information of the participant who trained the vertical federated learning model used by the user.

[0192] For example, the aforementioned client A and client B participate in the vertical federated task A. In the training phase, client A obtains a vertical federated learning model Θ A , and client B also obtains a vertical federated learning model Θ BSuppose client A and client B can both be participants in vertical federated task A and users of the vertical federated learning model. In the inference phase, if client A receives the identifier of vertical federated task A, client A can determine the corresponding vertical federated learning model Θ of vertical federated task A based on the identifier of vertical federated task A A , and then can calculate an intermediate result based on the vertical federated learning model Θ A Similarly, if client B receives the identifier of vertical federated task A, client B can determine the corresponding vertical federated learning model Θ of vertical federated task A based on the identifier of vertical federated task A B , and then can calculate an intermediate result based on the vertical federated learning model Θ B . In addition, client A can also determine the participant information of vertical federated task A based on the identifier of vertical federated task A. For example, the participant information of vertical federated task A includes the network entity identifier of client A, the role of client A in vertical federated task A is a subordinate participant, the network entity identifier of client B, and the role of client B in vertical federated task A is a primary participant. Then client A sends the intermediate result calculated by itself to client B. Alternatively, client B can also determine the above-mentioned participant information of vertical federated task A based on the identifier of vertical federated task A, and then client B can request the intermediate result from client A

[0193] It can be understood that client A and / or client B may also participate in other vertical federated tasks, that is, have multiple vertical federated learning models. Through the identifier of the vertical federated task, a vertical federated learning model can be uniquely determined, as well as the corresponding participant information of the vertical federated task and / or user information of the vertical federated learning model

[0194] Based on the above Figure 1 shown network system architecture and the content of the above related technical introduction, several possible communication methods are provided in the embodiments of the present application to implement the analysis task through the vertical federated learning model. In the following, the following methods can be implemented by the first network element or a module or chip in the first network element, or the second network element or a module or chip in the second network element. The first network element can be a core network element. For example, the first network element can be NWDAF (such as AnLF), AF or other core network elements. For example, the second network element can be NWDAF (such as MTLF) or other core network elements

[0195] As Figure 3 shown, the present application provides a communication method, and the method includes:

[0196] Step 300: The first network element sends a first message to the second network element. Correspondingly, the second network element receives the first message from the first network element.

[0197] Among them, the first message includes an analysis identifier, and the first message is used to request a model that provides services corresponding to the analysis identifier.

[0198] Exemplarily, the first message can also be referred to as a model request message. For example, the first message can be a model provision subscription (Nnwdaf_MLModelProvision_Subscribe) message.

[0199] In a possible implementation manner, before the first network element sends the first message to the second network element, the first network element can receive a fourth message from the service-consuming network element. The fourth message is used to request services corresponding to the analysis identifier, and the fourth message can include the analysis identifier. Among them, the service-consuming network element can also be referred to as the fourth network element, and the present application does not limit its name.

[0200] Step 310: The second network element sends a second message to the first network element. Correspondingly, the first network element receives the second message from the second network element.

[0201] Exemplarily, the second message can also be referred to as a model notification message. For example, the second message can be a model provision notification (Nnwdaf_MLModelProvision_Notify) message.

[0202] Among them, the second message includes one or more of a first identifier, information of M network elements, and information of a first model. Among them, the first identifier is associated with the analysis identifier. At least two vertically federated learning models corresponding to the first identifier, and the users of the at least two vertically federated learning models include M network elements. M is a positive integer. The at least two vertically federated learning models corresponding to the first identifier include the first model.

[0203] Exemplarily, the first identifier can be understood as an identifier of a vertically federated task, or the first identifier is used to identify a vertically federated task. Each vertically federated task is associated with at least two vertically federated learning models, or it can be understood that the at least two vertically federated learning models are obtained through the training stage of the vertically federated task. Therefore, at least two vertically federated learning models corresponding to the first identifier, or it can be understood that although the first identifier is the same, the vertically federated learning models identified by the first identifier are different among different participants in the vertically federated task.

[0204] Exemplarily, users of at least two vertical federated learning models can also be described as devices using at least two vertical federated learning models. It can also be understood as an entity that uses at least two vertical federated learning models to provide inference services (or inference results), or an entity that uses at least two vertical federated learning models to provide prediction services (or outputs), or an entity that uses at least two vertical federated learning models to provide analysis and identification corresponding services, or an entity that has the ability to use at least two vertical federated learning models.

[0205] Exemplarily, if the users of at least two vertical federated learning models include M network elements, then the M network elements can respectively use their corresponding vertical federated learning models to provide inference services. Among them, the M network elements correspond one-to-one with the M vertical federated learning models, and the at least two vertical federated learning models corresponding to the first identifier include the above-mentioned M vertical federated learning models.

[0206] Exemplarily, if the users of at least two vertical federated learning models include M network elements, then each of the M network elements can use at least one vertical federated learning model to provide inference services, that is, a network element can use one or more vertical federated learning models to provide inference services.

[0207] Exemplarily, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0208] For example, the information of the M network elements can be represented as a list of network element information, and the tuple in the list includes one or more of the network entity identifier, the subscription association address, and the role information.

[0209] Among them, taking the jth network element as an example, the jth network element is any one of the M network elements, and the role information of the jth network element indicates the role of the producer of the jth vertical federated learning model in the vertical federated task corresponding to the first identifier; for example, the role here is the main participant, or the subordinate participant, or the coordinator, etc.

[0210] Among them, the jth network element is the user of the jth vertical federated learning model, and the at least two vertical federated learning models corresponding to the first identifier include the jth vertical federated learning model.

[0211] Exemplarily, the number of vertical federated learning models corresponding to the first identifier is at least 2, and the first model is one of them. The first model is a vertical federated learning model, or the type of the first model is a vertical federated learning model. The second network element provides the information of the first model for the first network element. The first network element can be understood as the user of the first model, that is, the user of the vertical federated learning model. The second network element can be the producer of the first model and also a participant in the vertical federated task corresponding to the first identifier.

[0212] Exemplarily, the information of the first model may include the ML model file address (e.g., uniform resource locator (URL) or fully qualified domain name (FQDN)), or the analytics data repository functional (ADRF) ID or ADRF set ID. When including the ADRF ID or ADRF set ID, the first information may further include the storage transaction ID. Optionally, the information of the first model may include the identifier of the first model. In addition, the first information may further include other content, which is not limited in this application. It can be understood that part of the content included in the first information can be used to obtain the first model, and the first information further includes other content, such as information related to describing the first model, model accuracy information, model applicable scenario information, etc.

[0213] For example, if the information of the first model includes a uniform resource locator (URL), the first network element can obtain the first model according to this URL. Or, the information of the first model may include the ADRF ID, and the first network element can send a model acquisition request to the ADRF identified by this ADRF ID, and this ADRF can send the first model to the first network element.

[0214] Exemplarily, the information of the first identifier and / or the M network elements may be included in the information of the first model or located outside the information of the first model, which is not limited in this application. For example, the information of the first model includes the model file address, the first identifier and / or the information of the M network elements.

[0215] It can be understood that the information of the M network elements may be all user information or partial user information of at least two vertical federated learning models corresponding to the first identifier.

[0216] Example A, there are a total of M vertical federated learning models corresponding to the first identifier, that is, M≥2 at this time, and the information of the M network elements is all user information of at least two vertical federated learning models corresponding to the first identifier. That is to say, the M network elements may include the first network element.

[0217] In Example B, if there are M + 1 at least two vertical federated learning models corresponding to the first identifier, the information of M network elements is part of the user information of at least two vertical federated learning models corresponding to the first identifier, where the M network elements do not include the network element using the first model, that is, do not include the first network element. Since the first network element is the network element using the first model, it is not necessary to notify the information of the first network element again. That is to say, the M network elements may not include the first network element.

[0218] At this time, the first network element can obtain all the user information of at least two vertical federated learning models corresponding to the first identifier based on the information of the M network elements and its own information.

[0219] In addition, in a possible implementation, the second message further includes the information of N network elements, where the N network elements are participants in the vertical federated learning task corresponding to the first identifier, or producers of at least two vertical federated learning models corresponding to the first identifier, and N is a positive integer.

[0220] For example, the number of vertical federated learning models corresponding to the first identifier is 3, namely Model 1, Model 2, and Model 3. Among them, Network Element 1 trains Model 1, Network Element 2 trains Model 2, and Network Element 3 trains Model 3. It can be understood that the participants in the vertical federated learning task corresponding to the first identifier include Network Element 1, Network Element 2, and Network Element 3. For example, Network Element 1 and Network Element 2 can be secondary participants in the vertical federated learning task corresponding to the first identifier, and Network Element 3 can be the primary participant in the vertical federated learning task corresponding to the first identifier. It can also be described as that the producers of the 3 vertical federated learning models corresponding to the first identifier include Network Element 1, Network Element 2, and Network Element 3.

[0221] The information of the N network elements includes the information of Network Element 1, the information of Network Element 2, and the information of Network Element 3.

[0222] In an example, Network Element A is a user of Model 1, Network Element B is a user of Model 2, and Network Element C is a user of Model 3. That is, the users of the 3 vertical federated learning models corresponding to the first identifier include Network Element A, Network Element B, and Network Element C. The models used by Network Element A, Network Element B, and Network Element C are different.

[0223] Among them, Network Element 1, Network Element 2, Network Element 3, Network Element A, Network Element B, and Network Element C are all different. That is, the producer and user of Model 1 are different network elements, the producer and user of Model 2 are different network elements, and the producer and user of Model 3 are different network elements.

[0224] Combined with the above Example A, the information of the M network elements includes the information of Network Element A, the information of Network Element B, and the information of Network Element C.

[0225] Combined with the above Example B, if the first network element is Network Element A, the information of the M network elements includes the information of Network Element B and the information of Network Element C.

[0226] In another example, Network Element A is the user of Model 1, Network Element 2 is the user of Model 2, and Network Element 3 is the user of Model 3. That is, the users of the three vertical federated learning models corresponding to the first identifier include Network Element A, Network Element 2, and Network Element 3.

[0227] Among them, Network Element 1 is different from Network Element A. That is, the producer and user of Model 1 are different network elements, the producer and user of Model 2 are the same network element, and the producer and user of Model 3 are the same network element.

[0228] Combined with the above Example A, the information of the M network elements includes the information of Network Element A, the information of Network Element 2, and the information of Network Element 3.

[0229] Combined with the above Example B, if the first network element is Network Element A, the information of the M network elements includes the information of Network Element 2 and the information of Network Element 3.

[0230] The above examples are only for illustration, and there can be other examples, which are not limited in this application.

[0231] Exemplarily, the second network element can determine a vertical federated task or a vertical federated learning model according to the analysis identifier. Further, relevant information of the vertical federated task or the vertical federated learning model is obtained, specifically including: the first identifier, the information of the first model, and the information of the M network elements.

[0232] Exemplarily, the second network element can determine at least one of the first identifier, the first model, and the information of the M network elements in the following manner:

[0233] First, the second network element can determine at least one model according to the analysis identifier, and all of the at least one model can provide the service corresponding to the analysis identifier. Among them, the at least one model can include a general model (or a non-vertical federated learning model) and a vertical federated learning model, or all of the at least one model can be vertical federated learning models. Among them, the number of vertical federated learning models can also be one or more. In addition, if the second network element only determines a general model, the existing process can be referred to, which will not be elaborated here.

[0234] Among them, the specific process for the second network element to determine the vertical federated learning model can be as follows:

[0235] In a possible design, the second network element may determine a vertical federated learning model based on the analysis identifier. From the above relevant content about the vertical federated task, it can be known that the analysis identifier is related to the vertical federated task. For example, the analysis identifier only supports vertical federated tasks and does not support other machine learning tasks (such as the analysis of generating ordinary models). Further, the second network element may determine that the first model is a vertical federated learning model based on the analysis identifier.

[0236] In another possible design, the second network element may first determine a first identifier according to the analysis identifier. From the above relevant content about the vertical federated task, it can be known that the first identifier has an association relationship with the analysis identifier. Further, the second network element may determine one of the vertical federated learning models corresponding to the vertical federated task according to the first identifier, that is, the first model. It can be understood that the analysis identifier may have an association relationship with the identifiers of one or more vertical federated tasks, and then the second network element may determine one or more vertical federated learning models.

[0237] Then, the second network element may determine a model from at least one model by, but not limited to, the following methods.

[0238] Exemplarily, after a model is trained, the second network element may obtain performance information or performance parameters about the model, such as accuracy or error. In addition, the first message may also include information for indicating the model performance requirements. Further, the second network element may determine a model that meets the performance requirements from at least one model according to the information for indicating the model performance requirements in the first message and the performance information corresponding to each of the at least one model. Among them, the information for indicating the model performance requirements may be provided by the service consuming network element. For example, before step 300, the service consuming network element may also provide the information for indicating the model performance requirements to the first network element through the fourth message.

[0239] For example, the second network element may determine the general model A and the vertical federated learning model B according to the analysis identifier. Both the general model A and the vertical federated learning model B can provide the service corresponding to the analysis identifier. The first message further includes information for indicating the model performance requirements. Furthermore, the second network element may determine that the general model A does not meet the performance requirements and the vertical federated learning model B meets the performance requirements according to the performance information of the general model A, the performance information of the vertical federated learning model B, and the information for indicating the model performance requirements. That is, the second network element finally selects the vertical federated learning model B. Among them, the vertical federated learning model B can provide the service corresponding to the analysis identifier, which can be understood as that the vertical federated learning model B and other vertical federated learning models with the same vertical federated task identifier as the vertical federated learning model B can jointly provide the service corresponding to the analysis identifier. The vertical federated learning model B meets the performance requirements, that is, the vertical federated learning model B and other vertical federated learning models with the same vertical federated task identifier as the vertical federated learning model B jointly meet the performance requirements.

[0240] It can be understood that if at least one model includes a general model and a vertical federated learning model, the second network element may select the general model or the vertical federated learning model. Among them, if the general model is selected, the existing process can be referred to. The second network element provides the information of the general model for the first network element, and the first network element uses the general model to provide the service corresponding to the analysis identifier for the service-consuming network element, which will not be elaborated here. Only an example where the second network element selects a vertical federated learning model will be described below. Among them, the vertical federated learning model selected by the second network element is denoted as the first model.

[0241] Finally, after the second network element determines the first model, the second network element may further determine the information of M network elements and / or the information of N network elements. Exemplarily, the second network element has determined the first identifier in the process of determining the vertical federated learning model above. Furthermore, in combination with the relevant content about the vertical federated task above, it can be known that the second network element can also determine the participant information (that is, the information of N network elements) of the vertical federated task and / or the user information of the vertical federated learning model according to the first identifier. Among them, the information of M network elements can be part or all of the user information of the vertical federated learning model.

[0242] Exemplarily, after the training phase of the vertical federated task corresponding to the first identifier is completed, taking the i-th vertical federated learning model as an example, where the i-th vertical federated learning model is any one of at least two vertical federated learning models corresponding to the first identifier, the producer of the i-th vertical federated learning model or the user of the i-th vertical federated learning model may also send a notification message to the second network element, and the notification message includes information about the producer of the i-th vertical federated learning model and information about the user of the i-th vertical federated learning model. Among them, the producer of the i-th vertical federated learning model can also be understood as one of the participants in the vertical federated task corresponding to the first identifier.

[0243] That is, through the above process, the second network element can obtain the user information of at least two vertical federated learning models corresponding to the first identifier, and / or the producer information of at least two vertical federated learning models corresponding to the first identifier. Among them, the producer information of at least two vertical federated learning models corresponding to the first identifier can also be described as the participant information of the vertical federated task corresponding to the first identifier. In addition, the user information of at least two vertical federated learning models corresponding to the first identifier, and / or the producer information of at least two vertical federated learning models corresponding to the first identifier can also be saved in the second network element in a pre-configured manner.

[0244] In addition, in a possible implementation manner, the notification message may further include at least one of the information of the i-th vertical federated learning model. For example, the identifier of the i-th vertical federated learning model, etc. Alternatively, the second network element also saves the identifiers corresponding to at least two vertical federated learning models corresponding to the first identifier.

[0245] In a possible implementation manner, the first message may also be replaced with the information of M network elements and the identifiers of M models, where the M network elements and the M models are in one-to-one correspondence. The M models belong to at least two vertical federated learning models corresponding to the first identifier.

[0246] Exemplarily, taking the j-th network element as an example, where the j-th network element is any one of the M network elements and the j-th network element is the user of the j-th vertical federated learning model, the first network element may send the identifier of the j-th vertical federated learning model to the j-th network element so that the j-th network element calculates an intermediate result according to the j-th vertical federated learning model.

[0247] In another possible implementation manner, the first message may also be replaced with the information of M network elements and the identifiers of K models, where the M network elements and the K models have a corresponding relationship. The K models belong to at least two vertical federated learning models corresponding to the first identifier.

[0248] In a possible implementation, the second message may further include first indication information, where the first indication information indicates that the model providing the service corresponding to the analysis identifier is a vertical federated learning model. Alternatively, the first indication information indicates that the first model is a vertical federated learning model; or the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0249] Further, the first network element may determine that the type of the first model is a vertical federated learning model according to the first indication information. In addition, the first indication information is optional information. Since the second message includes information of M network elements, furthermore, the first network element may also determine that the type of the first model is a vertical federated learning model based on the information of the M network elements.

[0250] Further, optionally, step 320: The first network element sends the first identifier to at least one of the M network elements. Exemplarily, the first network element sends a third message to at least one of the M network elements, where the third message is used to request the vertical federated learning model corresponding to the first identifier to provide a service. For example, the first identifier may be carried by analyzing a subscription request message, or a (vertical federated learning) prediction request message, or a vertical federated learning model prediction request message. The network element that receives this message may determine the vertical federated learning model that it can use according to the first identifier, and then calculate an intermediate result according to the vertical federated learning model.

[0251] Exemplarily, the first network element determines at least one network element according to the information of the M network elements, and sends a third message to the at least one network element.

[0252] Exemplarily, in combination with the information of the M network elements, the cases where the first network element may determine at least one network element according to the information of the M network elements may include the following:

[0253] Combined with the above Example A, there are M vertical federated learning models corresponding to the first identifier, and the M network elements are all users of the vertical federated learning model corresponding to the first identifier. Then the first network element may determine the information of M - 1 network elements according to the information of the M network elements, where the M - 1 network elements do not include the first network element. Furthermore, the first network element may send a third message to each of the above M - 1 network elements.

[0254] Combined with the above Example B, there are M + 1 vertical federated learning models corresponding to the first identifier. Then the M network elements are partial users of the vertical federated learning model corresponding to the first identifier, where the M network elements do not include the first network element. Then the first network element may send a third message to each of the M network elements.

[0255] As can be seen from the above step 300, before the first network element sends the first message to the second network element, the first network element receives a fourth message from the service-consuming network element. The fourth message includes an analysis identifier and information about a first object, where the first object is the analysis object corresponding to the analysis identifier. The fourth message is used to request the service corresponding to the analysis identifier.

[0256] Exemplarily, the first object may also be referred to as a target UE (target UE). For example, the fourth message includes analysis identifier 1 and / or the SUPI of UE1. Here, analysis identifier 1 identifies the mobility service of the UE. If the first object is UE1, then the fourth message is used to request the analysis of the mobility of UE1.

[0257] The information about the first object includes the network entity identifier of the first object or the user entity identifier of the first object. Exemplarily, the user entity identifier is a subscription permanent identifier (SUPI).

[0258] In a possible implementation, the third message further includes the identifier of the first object in at least two vertical federated learning models corresponding to the first identifier. It can also be described as the identifier of the first object in the vertical federated task corresponding to the first identifier.

[0259] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is the mapping relationship between the first object and the object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier based on the identifier of the first object and the first mapping relationship.

[0260] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity identifier in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0261] In an example, the second message may further indicate or include a first mapping relationship, where the first mapping relationship is the identifier of the network entity identifier in at least two vertical federated learning models corresponding to the first identifier. Or it can be said to be the identifier of the network entity identifier in the vertical federated task corresponding to the first identifier. For example, the first mapping relationship can be implemented in the following way: (NFID, VFL ID).

[0262] Exemplarily, the first network element first obtains the network entity identifier of the first object. For example, the network entity identifier of the first object may be provided by the service consumption network element. For example, before step 300, the service consumption network element may also provide the network entity identifier of the first object to the first network element through the fourth message. Further, the first network element determines the identifier of the first object in at least two vertical federated learning models corresponding to the first identifier according to the network entity identifier of the first object and the first mapping relationship. At this time, the first object may be a network entity. For example, the first network element queries the first mapping relationship according to the NF ID of the first object to determine the VFL ID of the first object.

[0263] In another example, the second message may further indicate a second mapping relationship, where the second mapping relationship is the identifier of the user entity in at least two vertical federated learning models corresponding to the first identifier. For example, the second mapping relationship may be implemented in the following ways: (SUPI, VFL ID), or (UE ID, UE VFL ID).

[0264] Exemplarily, the first network element first obtains the user entity identifier of the first object. For example, the user entity identifier of the first object may be provided by the service consumption network element. For example, before step 300, the service consumption network element may also provide the user entity identifier of the first object to the first network element through the fourth message. Further, the first network element determines the identifier of the first object in the vertical federated learning model according to the user entity identifier of the first object and the second mapping relationship. At this time, the first object may be a terminal device, such as a UE. For example, the first network element queries the second mapping relationship according to the SUPI of the first object to determine the VFL ID of the first object. Or, the first network element queries the second mapping relationship according to the UE ID of the first object to determine the UE VFL ID of the first object.

[0265] In another possible implementation, the third message further includes information about the first object.

[0266] For example, the network entity identifier of the first object or the user entity identifier of the first object. If the third message further includes the network entity identifier of the first object, the network element that receives the network entity identifier of the first object may determine the identifier of the first object in the vertical federated learning model according to the first mapping relationship and the network entity identifier of the first object. Or, if the third message further includes the user entity identifier of the first object, the network element that receives the user entity identifier of the first object may determine the identifier of the first object in the vertical federated learning model according to the second mapping relationship and the user entity identifier of the first object.

[0267] Further, after the first network element sends a third message to at least one network element, in a possible implementation, the first network element obtains a first intermediate result according to a first model; the first network element receives at least one intermediate result from at least one network element, where the at least one network element corresponds to the at least one intermediate result one by one; the first network element determines an output according to the first intermediate result and the at least one intermediate result. In another possible implementation, the first network element obtains a first intermediate result according to a first model; the first network element sends the first intermediate result to a third network element; the third network element is one of the at least one network element; the first network element receives an output from the third network element.

[0268] Next, according to the different roles of the producer of the first model, the first network element can obtain an output through the following several possible ways.

[0269] In the following text, the producer of the first model is the main participant or the secondary participant in the vertical federated task corresponding to the first identifier. For the sake of convenient description, hereinafter it is simply referred to as the producer of the first model being the main participant or the secondary participant. Among them, the producer of the first model being the main participant (or the secondary participant) can also be described as the model obtained by the first network element (i.e., the first model) being the model generated by the main participant (or the secondary participant) in the vertical federated task corresponding to the first identifier, or the second network element being the main participant (or the secondary participant) in the vertical federated task corresponding to the first identifier. The following takes the third message including the first identifier and the identifier of the first object in at least two vertical federated learning models corresponding to the first identifier (hereinafter simply referred to as the VFL ID of the first object) as an example for illustration.

[0270] Method 1: The producer of the first model is the main participant.

[0271] Exemplarily, in combination with the above Example A, the information of the M network elements includes the information of the first network element and the information of the other M - 1 network elements. Among them, the information of the first network element includes the role information of the first network element, and the role information of the first network element indicates that the producer of the first model is the main participant. Or, in combination with the above Example B, the information of the M network elements does not include the information of the first network element, the information of the M network elements includes the role information of the M network elements, the role information of the M network elements can be one piece of information or M pieces of information, and the role information of the M network elements indicates that the producers of the models used by the M network elements are all secondary participants in the vertical federated task corresponding to the first identifier. Hereinafter simply referred to as the producers of the models used by the M network elements are all secondary participants, then the first network element determines that the producer of the first model is the main participant according to the role information of the M network elements.

[0272] Exemplarily, the first network element determines at least one network element according to the information of M network elements, and sends a third message to each of the at least one network element. The third message includes a first identifier and the identifier of the first object in at least two vertical federated learning models corresponding to the first identifier, hereinafter simply referred to as the VFL ID of the first object. Exemplarily, in combination with the above example A, the number of at least one network element is M - 1, and in combination with the above example B, the number of at least one network element is M. The network element that receives the first identifier and the VFL ID of the first object can determine the corresponding vertical federated learning model according to the first identifier, and calculate the corresponding intermediate result according to the VFL ID of the first object and the determined vertical federated learning model. Wherein, each of the at least one network element determines an intermediate result. The above at least one network element sends the corresponding intermediate result to the first network element respectively. That is, the first network element receives at least one intermediate result from at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one. The first network element also obtains a first intermediate result according to the first model. Further, the first network element determines the output according to the first intermediate result and the at least one intermediate result received. The following combines Figure 4 to illustrate the above process with an example.

[0273] Figure 4 In this case, the number of at least one network element is 2, and the at least one network element is network element X and network element Y respectively. Network element X is the user of vertical federated learning model X, and network element Y is the user of vertical federated learning model Y. The producers of both vertical federated learning model X and vertical federated learning model Y are subordinate participants.

[0274] S401: The first network element sends the first identifier and the VFL ID of the first object to network element X and network element Y respectively.

[0275] S402A: Network element X determines vertical federated learning model X according to the first identifier, and determines intermediate result X according to vertical federated learning model X and the VFL ID of the first object.

[0276] S402B: Network element Y determines vertical federated learning model Y according to the first identifier, and determines intermediate result Y according to vertical federated learning model Y and the VFL ID of the first object.

[0277] S403A: Network element X sends intermediate result X to the first network element.

[0278] S403B: Network element Y sends intermediate result Y to the first network element.

[0279] S404: The first network element obtains a first intermediate result according to the first model.

[0280] Exemplarily, the first network element calculates a first intermediate result based on the VFL ID of the first object and the first model. This application does not limit the order of S404 with respect to S401, S402A, S402B, S403A, and S403B.

[0281] S405: The first network element determines an output based on the first intermediate result, intermediate result X, and intermediate result Y.

[0282] Mode 2: The producer of the first model is a subordinate participant.

[0283] Among them, the producer of the first model is a subordinate participant, and the producer of the model used by the third network element is a primary participant.

[0284] Exemplarily, in combination with the above example A, if M = 2, the information of the M network elements includes the information of the first network element and the information of the third network element. If M > 2, the information of the M network elements includes the information of the first network element, the information of the third network element, and the information of the other M - 2 network elements. Among them, the information of the first network element includes the role information of the first network element, and the role information of the first network element indicates that the producer of the first model is a subordinate participant. The information of the third network element includes the role information of the third network element, and the role information of the third network element indicates that the producer of the model used by the third network element is a primary participant. In combination with the above example B, the information of the M network elements does not include the information of the first network element. If M = 2, the information of the M network elements includes the information of the third network element. If M > 2, the information of the M network elements includes the information of the third network element and the information of the other M - 2 network elements. Among them, the information of the third network element includes the role information of the third network element, and the role information of the third network element indicates that the producer of the model used by the third network element is a primary participant. Furthermore, the first network element can determine that the producer of the first model is a subordinate participant based on the information of the third network element.

[0285] Exemplarily, the first network element determines at least one network element based on the information of the M network elements, and sends a third message to each of the at least one network element. The third message includes a first identifier and the VFL ID of the first object. Exemplarily, in combination with the above example A, the number of the at least one network element is M - 1. In combination with the above example B, the number of the at least one network element is M. The network element that receives the first identifier and the VFL ID of the first object can determine the corresponding vertical federated learning model based on the first identifier, and calculate an intermediate result based on the VFL ID of the first object and the determined vertical federated learning model. Among them, each network element in the at least one network element determines an intermediate result.

[0286] Next, the following two possible implementation manners may be included:

[0287] Possible implementation a: The network elements other than the third network element among the above at least one network element send corresponding intermediate results to the first network element. That is, the first network element receives the intermediate results from the network elements other than the third network element among the at least one network element. The first network element further obtains a first intermediate result according to the first model, and further sends the first intermediate result and the received intermediate results to the third network element. The third network element can determine the output according to the intermediate results received from the first network element and the intermediate results calculated by itself, and send the output to the first network element.

[0288] Possible implementation b: The network elements other than the third network element among the above at least one network element send corresponding intermediate results to the third network element. The first network element further obtains a first intermediate result according to the first model, and further sends the first intermediate result to the third network element. The third network element can determine the output according to the intermediate results from the network elements other than the third network element among the at least one network element, the first intermediate result, and the intermediate results calculated by itself, and send the output to the first network element.

[0289] The following combines Figure 5 to illustrate the above process by way of example.

[0290] Figure 5 Among them, the number of at least one network element is 2, and the at least one network element is network element X and network element Y respectively. Among them, network element X is the user of the vertical federated learning model X, network element Y is the user of the vertical federated learning model Y, the producer of the vertical federated learning model X is the main participant, and the producer of the vertical federated learning model Y is the subordinate participant.

[0291] S501: The first network element sends the first identifier and the VFL ID of the first object to network element X and network element X respectively.

[0292] S502A: Network element X determines the vertical federated learning model X according to the first identifier, and determines the intermediate result X according to the vertical federated learning model X and the VFL ID of the first object.

[0293] S502B: Network element Y determines the vertical federated learning model Y according to the first identifier, and determines the intermediate result Y according to the vertical federated learning model Y and the VFL ID of the first object.

[0294] S503: Network element Y sends the intermediate result Y to the first network element.

[0295] S504: The first network element obtains the first intermediate result according to the first model.

[0296] This application does not limit the order of S504 and S501, S502A, S502B, S503.

[0297] S505: The first network element sends the first intermediate result and the intermediate result Y to network element X.

[0298] Alternatively, S503 can be replaced by the network element Y sending the intermediate result Y to the network element X, and at the same time, S505 can be replaced by the first network element sending the first intermediate result to the network element X.

[0299] S506: The network element X determines the output based on the first intermediate result, the intermediate result X, and the intermediate result Y.

[0300] S507: The network element X sends the output to the first network element.

[0301] As an alternative embodiment, the number of vertically federated learning models corresponding to the first identifier is greater than 2, the producer of the first model is a subordinate participant, the information of the M network elements only includes the information of the third network element, that is, M = 1. The information of the third network element includes the role information of the third network element, and the role information of the third network element indicates that the producer of the model used by the third network element is the main participant. That is, in addition to the first network element and the third network element, there is at least one other network element, and the producers of the models used by the at least one other network element are all subordinate participants. Among them, the first network element can determine that the producer of the first model is a subordinate participant according to the information of the third network element. At this time, the third network element only knows its own information and the information of the third network element, but the first network element does not obtain the information of the at least one other network element. Then the third network element can initiate the collection of the intermediate results of the at least one other network element and provide the output. The following is combined with Figure 6 An example is given to illustrate the above process.

[0302] Figure 6 In [description], the number of vertically federated learning models corresponding to the first identifier = 3. Among them, the network element X is the user of the vertically federated learning model X, the network element Y is the user of the vertically federated learning model Y, the producer of the vertically federated learning model X is the main participant, and the producer of the vertically federated learning model Y is the subordinate participant.

[0303] S601: The first network element sends the first identifier and the VFL ID of the first object to the network element X.

[0304] S602: The first network element obtains the first intermediate result according to the first model.

[0305] S603: The first network element sends the first intermediate result to the network element X.

[0306] It can be understood that the first network element can also carry the first intermediate result, the first identifier, and the VFL ID of the first object in one message.

[0307] Among them, S602 and S603 can be before or after S601, and this application does not make a limitation on this.

[0308] S604: The network element X determines the vertical federated learning model X according to the first identifier, and determines the intermediate result X according to the vertical federated learning model X and the VFL ID of the first object.

[0309] S605: The network element X determines the network element that has not obtained the intermediate result of the network element Y, and sends the first identifier and the VFL ID of the first object to the network element Y.

[0310] Exemplarily, the network element X determines the information of all users of the vertical federated learning model corresponding to the first identifier according to the first identifier. All users of the vertical federated learning model corresponding to the first identifier include the first network element, the network element X, and the network element Y. Furthermore, the network element X determines the network element that has not obtained the intermediate result of the network element Y according to the obtained intermediate result.

[0311] S606: The network element Y determines the vertical federated learning model Y according to the first identifier, and determines the intermediate result Y according to the vertical federated learning model Y and the VFL ID of the first object.

[0312] S607: The network element Y sends the intermediate result Y to the network element X.

[0313] S608: The network element X determines the output according to the first intermediate result, the intermediate result X, and the intermediate result Y.

[0314] S609: The network element X sends the output to the first network element.

[0315] In addition, for the above-mentioned method 1 and method 2, the network element using the producer of the model as the main participant can also determine the information of all users of the vertical federated learning model corresponding to the first identifier according to the first identifier, and judge whether all intermediate results have been obtained according to the obtained intermediate results, which will not be elaborated here.

[0316] As Figure 7 shown, the present application also provides a communication method, which includes:

[0317] Step 700: The first network element sends a first message to the second network element. Correspondingly, the second network element receives the first message from the first network element.

[0318] Among them, the first message includes an analysis identifier, and the first message is used to request a model for providing services corresponding to the analysis identifier.

[0319] Exemplarily, the first message can also be called a model request message.

[0320] In a possible implementation manner, before the first network element sends the first message to the second network element, the first network element can receive a fourth message from the service-consuming network element. The fourth message is used to request services corresponding to the analysis identifier, and the fourth message may include the analysis identifier. Other relevant content about the fourth message can be referred to the above Figure 3The illustrated embodiment.

[0321] Step 710: The second network element sends a second message to the first network element. Correspondingly, the first network element receives the second message from the second network element.

[0322] Exemplarily, the second message can also be referred to as a model notification message.

[0323] Wherein, the second message includes a second identifier and / or first indication information. The second identifier is used to identify the first message, and the first indication information indicates that the model providing the service corresponding to the analysis identifier is a vertical federated learning model. Further, the first network element can determine that the model providing the service corresponding to the analysis identifier is a vertical federated learning model according to the first indication information.

[0324] Exemplarily, the second identifier can be a subscription correlation ID. The second identifier is associated with the vertical federated learning model subscription and is used to identify the model subscription between the second network element and the first network element. In addition, the second network element saves the corresponding relationship between the second identifier and the vertical federated learning model.

[0325] Exemplarily, before the second network element sends the second message to the first network element, that is, before step 710, the second network element can determine to use the first model to provide the service corresponding to the analysis identifier, and the first model is a vertical federated learning model. Furthermore, the second network element saves the information corresponding relationship between the second identifier and the first model, so that after the second network element receives the second identifier, it can determine the information of the first model according to the second identifier. The information about the first model can refer to Figure 3 the relevant descriptions in the illustrated embodiment.

[0326] When the second network element determines the information of the first model according to the analysis identifier, the second network element can also determine the information of M network elements and / or the first identifier according to the analysis identifier;

[0327] When the second network element saves the corresponding relationship between the second identifier and the information of the first model, the second network element saves the corresponding relationship between the second identifier and the information of the first model, the information of M network elements and the first identifier; or the corresponding relationship between the second identifier and the information of the first model, the information of M network elements; or the corresponding relationship between the second identifier and the information of the first model, the first identifier, etc.

[0328] Wherein, the first identifier is associated with the analysis identifier, and there are at least two vertical federated learning models corresponding to the first identifier; the users of the at least two vertical federated learning models include M network elements, M is a positive integer, and the at least two vertical federated learning models include the first model. Specifically, reference can be made to the relevant descriptions in the above Figure 3 illustrated embodiment.

[0329] Among them, for the specific process of the second network element to determine the first model, reference can be made to the relevant descriptions of the second network element determining at least one of the first identifier, the first model, and the information of M network elements in the embodiments shown above, which will not be elaborated here. Figure 3 shown in the embodiments above, which will not be elaborated here.

[0330] Step 720: The first network element sends a third message to the second network element. The third message is used to request an analysis of the service corresponding to the identifier, and the third message includes a second identifier.

[0331] Exemplarily, the third message may further include the identifier of the first object in at least two vertical federated learning models corresponding to the first identifier, or the information of the first object, for example, the network entity identifier or user entity identifier of the first object. The specific possible implementation manners can refer to the relevant content in the embodiments shown above, which will not be elaborated here. Figure 3 shown in the embodiments above, which will not be elaborated here.

[0332] Exemplarily, the second network element determines the information of the first model according to the second identifier. The second network element uses the first model and provides an analysis of the service corresponding to the identifier to the first network element.

[0333] Therefore, in this embodiment, the second network element does not provide the information of the model to the first network element, but uses the first indication information to notify the first network element that the model for analyzing the service corresponding to the identifier is a vertical federated learning model, and uses the second identifier to enable the second network element to request the service corresponding to the identifier.

[0334] Furthermore, the first network element may also receive the output from the second network element, and the first network element may also send the output to the service consumption network element.

[0335] Exemplarily, if the second network element stores the correspondence between the second identifier, the information of the first model, the information of M network elements, and the first identifier, after the second network element receives the third message, the second network element may determine at least one of the information of the first model, the information of M network elements, and the first identifier according to the second identifier in the third message. Further, in combination with at least one of the information of the first model, the information of M network elements, and the first identifier, the second network element may also obtain an output.

[0336] In a possible design, after the second network element receives a third message from the first network element, the second network element determines information about the first model, information about M network elements, and a first identifier according to a second identifier. The second network element determines at least one network element according to the information about the M network elements, and the second network element sends a fifth message to the at least one network element. The fifth message includes the first identifier, and the fifth message is used to request the vertical federated learning model corresponding to the first identifier to provide services. The second network element obtains a first intermediate result according to the first model and receives at least one intermediate result from the at least one network element, and there is a one-to-one correspondence between the at least one network element and the at least one intermediate result. The second network element determines an output according to the first intermediate result and the at least one intermediate result.

[0337] In a possible design, after the second network element receives a third message from the first network element, the second network element determines information about the first model, information about M network elements, and a first identifier according to a second identifier. The second network element determines at least one network element according to the information about the M network elements and sends a fifth message to the at least one network element. The fifth message includes the first identifier, and the third message is used to request the vertical federated learning model corresponding to the first identifier to provide services. The second network element obtains a first intermediate result according to the first model, sends the first intermediate result to the third network element, and receives an output from the third network element.

[0338] Among them, the fifth message may further include an identifier of the first object in at least two vertical federated learning models corresponding to the first identifier, or information about the first object. For example, the network entity identifier or user entity identifier of the first object. The specific possible implementation manners may refer to the relevant content in the above Figure 3 illustrated embodiments, which will not be elaborated here. The following takes the fifth message including the first identifier and the VFL ID of the first object as an example for illustration.

[0339] Exemplarily, according to different roles of the second network element, the second network element may obtain an output in the following two ways: Way A and Way B:

[0340] Way A: The role of the second network element is the main participant.

[0341] Exemplarily, the second network element may learn the information about the M network elements. Specifically, the second network element may determine all the user information of the vertical federated learning model corresponding to the first identifier. Therefore, the second network element may learn its own role and the roles of the producers of the models used by other users in the vertical federated task corresponding to the first identifier. The second network element may determine the information about at least one network element according to the information about the M network elements. Exemplarily, the information about at least one network element is the user information of all the users of the vertical federated learning model corresponding to the first identifier except the second network element.

[0342] Exemplarily, the second network element sends a fifth message to at least one network element respectively. The fifth message includes a first identifier and the VFL ID of a first object. The network element that receives the first identifier and the VFL ID of the first object can determine the corresponding vertical federated learning model according to the first identifier, and calculate the corresponding intermediate result according to the VFL ID of the first object and the determined vertical federated learning model. Each of the at least one network elements determines an intermediate result. The above at least one network element sends the corresponding intermediate result to the second network element respectively. That is, the second network element receives at least one intermediate result from at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one. The second network element also obtains a first intermediate result according to the first model. Further, the second network element determines the output according to the first intermediate result and the at least one intermediate result received.

[0343] The above Method A can refer to Figure 4 the embodiments shown, where Figure 4 the first network element in the embodiments shown can be replaced by the second network element.

[0344] Method B: The role of the second network element is a slave participant.

[0345] Among them, the role of the second network element is a slave participant, and the role of the third network element using the role of the model producer is a master participant.

[0346] Exemplarily, the second network element can obtain the information of M network elements. Specifically, the second network element can determine all the user information of the vertical federated learning model corresponding to the first identifier. Therefore, the second network element can know its own role, and the role of the model producers used by other users in the vertical federation task corresponding to the first identifier. The second network element can determine the information of at least one network element according to the information of the M network elements. Exemplarily, the information of at least one network element is the user information of the users other than the second network element among all the users of the vertical federated learning model corresponding to the first identifier.

[0347] Exemplarily, the second network element sends the first identifier and the VFL ID of the first object to at least one network element respectively. The network element that receives the first identifier and the VFL ID of the first object can determine the corresponding vertical federated learning model according to the first identifier, and calculate the intermediate result according to the VFL ID of the first object and the determined vertical federated learning model. Each network element in the at least one network element determines an intermediate result. The network elements other than the third network element in the at least one network element send the corresponding intermediate results to the second network element. That is, the second network element receives the intermediate results from the network elements other than the third network element in the at least one network element. The second network element also obtains the first intermediate result according to the first model, and further sends the first intermediate result and the received intermediate results to the third network element. The third network element can determine the output according to the intermediate results received from the second network element and the intermediate results calculated by itself, and send the output to the second network element.

[0348] The above method B can refer to Figure 5 the embodiment shown, where Figure 5 the first network element in the embodiment shown can be replaced by the second network element.

[0349] In addition, in a possible implementation, the second message includes a model subscription address and first indication information. The first network element can send a third message to this address, which is an address of the second network element, and the second network element stores the corresponding relationship between this address and the information of the first model. The second network element determines the information of the first model according to the address in the third message. The second network element uses the first model and provides the service corresponding to the analysis identifier to the first network element. Exemplarily, the model subscription address is similar in function to the second identifier and can refer to the above related description.

[0350] In the following, the first network element is taken as the first AnLF and the second network element is taken as the first MTLF as an example for illustration.

[0351] Figure 8 The embodiment shown is for the above Figure 3 shown embodiment for further illustration.

[0352] S801: The service consumption network element sends a first analysis subscription message to the first AnLF. The first analysis subscription message includes an analysis identifier, the SUPI or NF ID of object X. Optionally, the first analysis subscription message further includes information for indicating the model performance requirements.

[0353] Exemplarily, the service consumption network element requests the first AnLF to provide the service corresponding to the analysis identifier through the first analysis subscription message. For example, the service consumption network element can be an AMF, an SMF or other core network elements.

[0354] Exemplarily, the first analysis subscription message can specifically refer to the relevant description of the fourth message above.

[0355] S802: The first AnLF sends a first message to the first MTLF. The first message includes an analysis identifier.

[0356] In a possible implementation, the first AnLF can determine the first MTLF through a network element discovery process. Exemplarily, the first AnLF sends an NF discovery message to the NRF. The NF discovery message includes an analysis identifier. For example, the NF discovery message can be Nnrf_NF Discovery. The NF discovery message is used to discover the MTLF or the NWDAF that contains the MTLF. The NRF determines the first MTLF according to the analysis identifier in the NF discovery message. The NRF sends an NF discovery response message to the first AnLF. The NF discovery response message includes information about the first MTLF. The information about the first MTLF can include the identifier of the first MTLF, the analysis identifier, etc. For example, the information about the first MTLF is the NF information (profile) of the first MTLF.

[0357] It can be understood that the NRF can determine one or more MTLFs according to the analysis identifier in the NF discovery message. At this time, the NF discovery response message can include information about the one or more MTLFs. The first MTLF can be understood as any one of the one or more MTLFs. Hereinafter, only the first MTLF is taken as an example for illustration.

[0358] S803: The first MTLF sends a second message to the first AnLF.

[0359] The second message includes at least one of the information of the first model, the information of M network elements, and the first identifier.

[0360] Exemplarily, the first MTLF determines at least one of the information of the first model, the information of M network elements, and the first identifier according to the analysis identifier in the model request message. Specifically, it can refer to the relevant content in the Figure 3 illustrated embodiments above, which will not be elaborated here.

[0361] Optionally, the second message can further include a first mapping relationship and / or a second mapping relationship.

[0362] S804: The first AnLF determines at least one network element according to the information of M network elements.

[0363] S805: The first AnLF sends the first identifier and the VFL ID of object X to each of the at least one network element.

[0364] Optionally, the first AnLF determines the VFL ID of object X according to the SUPI or NF ID of object X in the first analysis subscription message, and the first mapping relationship and / or the second mapping relationship included in the second message.

[0365] Hereinafter, only the role of the first model producer as the main participant will be taken as an example for illustration, which is not a limitation of this application.

[0366] S806: At least one network element sends at least one intermediate result to the first AnLF, and the at least one intermediate result corresponds to the at least one network element one by one.

[0367] S807: The first AnLF obtains the first intermediate result according to the first model.

[0368] Exemplarily, the first AnLF calculates the first intermediate result according to the VFL ID of object X and the first model.

[0369] S808: The first AnLF determines the output according to the first intermediate result and the at least one intermediate result.

[0370] S809: The first AnLF sends the output to the service consumption network element.

[0371] By adopting the above method, the first AnLF can obtain at least one of the information of the first model, the information of the M network elements, and the first identifier from the first MTLF, and perform an analysis task by using the vertical federated learning model together with at least one of the M network elements.

[0372] Figure 9 The illustrated embodiment is directed to the above Figure 7 The illustrated embodiment is further illustrated by way of example.

[0373] S901 to S902 may refer to the above S801 to S802.

[0374] S903: The first MTLF sends a second message to the first AnLF.

[0375] Wherein, the second message includes a second identifier and / or first indication information.

[0376] Exemplarily, the first MTLF determines at least one of the information of the first model, the information of the M network elements, and the first identifier according to the analysis identifier in the model request message. Specifically, reference may be made to the relevant content in the above Figure 3 illustrated embodiment. The first MTLF stores the corresponding relationship between the second identifier and the information of the first model.

[0377] Optionally, the second message may further include a first mapping relationship and / or a second mapping relationship.

[0378] S904: The first AnLF sends a third message to the first MTLF.

[0379] Among them, the third message includes the VFL ID of object X and the second identifier.

[0380] Optionally, the first AnLF determines the VFL ID of object X according to the SUPI or NF ID of object X in the first analysis subscription message, and the first mapping relationship and / or the second mapping relationship included in the second message.

[0381] S905: The first MTLF determines at least one of the information of the first model, the information of M network elements, and the first identifier according to the second identifier.

[0382] S906: The first MTLF determines at least one network element according to the information of M network elements.

[0383] S907: The first MTLF sends the first identifier and the VFL ID of object X to at least one network element respectively.

[0384] The following only takes the role of the first MTLF as the main participant as an example for illustration, and is not a limitation of this application.

[0385] S908: At least one network element sends at least one intermediate result to the first MTLF, and at least one intermediate result corresponds to at least one network element one by one.

[0386] S909: The first MTLF obtains a first intermediate result according to the first model.

[0387] Exemplarily, the first MTLF calculates the first intermediate result according to the VFL ID of object X and the first model.

[0388] S910: The first MTLF determines the output according to the first intermediate result and at least one intermediate result.

[0389] S911: The first MTLF sends the output to the first AnLF.

[0390] S912: The first AnLF sends the output to the service consumption network element.

[0391] By adopting the above method, the first AnLF does not obtain the information of the model from the first MTLF, but can determine to use the vertical federated learning model through the obtained first indication information, and request the service corresponding to the analysis identifier through the second identifier. The first MTLF and at least one network element among the M network elements jointly execute the analysis task by adopting the vertical federated learning model.

[0392] This application also provides a communication method, as Figure 10 shown, this method includes:

[0393] S1001: The service consumption network element sends a first analysis subscription message to the first AnLF. The first analysis subscription message includes an analysis identifier, the SUPI or NF ID of object X, and information indicating model performance requirements.

[0394] S1002: The first AnLF determines that it cannot obtain a model that meets the performance requirements.

[0395] Exemplarily, the first AnLF may request to obtain a model from the model production network element, but the first AnLF cannot obtain a model that meets the model performance requirements. For example, the model production network element (such as MTLF) cannot generate a model that meets the model performance requirements and sends a model request rejection message. The first AnLF determines that it cannot obtain a model that meets the requirements based on the model request rejection message.

[0396] For example, the first AnLF may obtain a general model and determine that the performance of the general model does not meet the performance requirements.

[0397] S1003: The first AnLF sends a service rejection message to the service consumption network element.

[0398] The service rejection message indicates that the first AnLF cannot obtain a model that meets the performance requirements.

[0399] S1004: The service consumption network element sends an NF discovery message to the NRF.

[0400] The NF discovery message is used to discover the NWDAF containing MTLF and AnLF. The NF discovery message includes an analysis identifier and third indication information, and the third indication information is used to request a NWDAF with vertical federated learning capabilities. Optionally, the NF discovery message further includes fifth indication information, which is used to instruct the NRF to return the NWDAF containing MTLF and AnLF. The fifth indication information and the third indication information may also be combined into one indication information.

[0401] S1005: The NRF sends an NF discovery response message to the service consumption network element.

[0402] Exemplarily, the NRF determines a first NWDAF based on the NF discovery message. The first NWDAF includes MTLF and AnLF, and the first NWDAF has the ability for vertical federated learning. The NF discovery response message includes information about the first NWDAF. Among them, when the first NWDAF sends a registration request message to the NRF, the registration request message indicates that the first NWDAF includes MTLF and AnLF, and the first NWDAF has the ability for vertical federated learning. The NRF determines the NWDAFs that support both MTLF and AnLF according to the fifth indication information, determines the NWDAFs with the ability for vertical federated learning according to the third indication information, and determines the NWDAFs that support both MTLF and AnLF and have the ability for vertical federated learning according to the fifth indication information and the third indication information.

[0403] S1006: The service consumption network element sends a first message to the first NWDAF. Among them, the first message includes an analysis identifier and a fourth indication information, and the fourth indication information indicates that the first NWDAF generates a vertical federated learning model corresponding to the analysis identifier. The first message also includes information for indicating the model performance requirements.

[0404] S1007: The first NWDAF initiates a vertical federated task W and generates a first model, and the first model is a vertical federated learning model.

[0405] Exemplarily, the first NWDAF participates in the vertical federated task W together with at least one other network element. The vertical federated task W is used to train a vertical federated learning model corresponding to the analysis identifier. The first NWDAF obtains the first model, the first model is a vertical federated learning model, the first model meets the performance requirements, and the first model is used to provide services corresponding to the analysis identifier. The first NWDAF saves the correspondence between the second identifier and the information of the first model. The second identifier is used to identify the first message.

[0406] S1008: The first NWDAF sends a second message to the service consumption network element. Among them, the second message includes the second identifier and the first indication information.

[0407] Optionally, the first NWDAF also sends the first mapping relationship and / or the second mapping relationship to the service consumption network element.

[0408] S1009: The service consumption network element sends a third message to the first NWDAF. The third message includes the second identifier and the VFL ID of object X.

[0409] Optionally, the service consumption network element determines the VFL ID of object X according to the SUPI or NF ID of object X, and the first mapping relationship and / or the second mapping relationship.

[0410] S1010: The first NWDAF determines information of the first model, information of at least one other network element, and an identifier of the vertical federated task W according to a second identifier.

[0411] Exemplarily, the first NWDAF stores a correspondence relationship between the second identifier and information of the first model, information of at least one other network element, and an identifier of the vertical federated task W.

[0412] S1011: The first NWDAF separately sends the identifier of the vertical federated task W and the VFL ID of the object X to at least one other network element.

[0413] The following is only illustrated by taking the role of the first NWDAF as the main participant, which is not a limitation of this application.

[0414] S1012: At least one other network element sends at least one intermediate result to the first NWDAF, and at least one intermediate result corresponds to at least one other network element one by one.

[0415] S1013: The first NWDAF obtains a first intermediate result according to the first model.

[0416] Exemplarily, the first NWDAF calculates the first intermediate result according to the VFL ID of the object X and the first model.

[0417] S1014: The first NWDAF determines an output according to the first intermediate result and at least one intermediate result.

[0418] S1015: The first NWDAF sends the output to the service consumption network element.

[0419] By using the above method, the service consumption network element can discover an NWDAF that supports MTLF and AnLF and has the ability of vertical federated learning, and instruct the NWDAF to initiate a vertical federated task, and provide services corresponding to the analysis identifier through the vertical federated learning model obtained by the vertical federated task.

[0420] This application also provides a communication method, as Figure 11 shown, the method includes:

[0421] S1101: The service consumption network element sends a first analysis subscription message to the first AnLF. The first analysis subscription message includes an analysis identifier, the SUPI or NF ID of the object X, and information for indicating model performance requirements.

[0422] S1102: The first AnLF determines that a model meeting the performance requirements cannot be obtained.

[0423] S1103: The first AnLF sends an NF discovery message to the NRF.

[0424] The NF discovery message is used to discover the NWDAF that includes MTLF and AnLF. The NF discovery message includes an analysis identifier and third indication information, and the third indication information is used to request the NWDAF with the ability of vertical federated learning. Optionally, the NF discovery message further includes fifth indication information, and this indication information is used to instruct the NRF to return the NWDAF that includes MTLF and AnLF. Among them, the fifth indication information and the third indication information can also be combined into one indication information.

[0425] S1104: The first AnLF sends an NF discovery response message to the NRF.

[0426] Exemplarily, the NRF determines the first NWDAF according to the NF discovery message. The first NWDAF includes MTLF and AnLF, and the first NWDAF has the ability of vertical federated learning. The NF discovery response message includes the information of the first NWDAF. Among them, when the first NWDAF sends a registration request message to the NRF, the registration request message indicates that the first NWDAF includes MTLF and AnLF, and the first NWDAF has the ability of vertical federated learning. The NRF determines the NWDAF that supports both MTLF and AnLF according to the fifth indication information, determines the NWDAF with the ability of vertical federated learning according to the third indication information, and determines the NWDAF that supports MTLF and AnLF and has the ability of vertical federated learning according to the fifth indication information and the third indication information.

[0427] S1105: The first AnLF sends a first message to the first NWDAF. Among them, the first message includes an analysis identifier and fourth indication information, and the fourth indication information instructs the first NWDAF to generate a vertical federated learning model corresponding to the analysis identifier. The first message further includes information used to indicate the model performance requirements.

[0428] S1106: The first NWDAF initiates a vertical federated task W and generates a first model, and the first model is a vertical federated learning model.

[0429] Exemplarily, the first NWDAF participates in the vertical federated task W together with at least one other network element. The vertical federated task W is used to train the vertical federated learning model corresponding to the analysis identifier. The first NWDAF obtains the first model, the first model is a vertical federated learning model, the first model meets the performance requirements, and the first model is used to provide the service corresponding to the analysis identifier. The first NWDAF saves the corresponding relationship between the second identifier and the information of the first model. The second identifier is used to identify the first message.

[0430] S1107: The first NWDAF sends a second message to the first AnLF, and the second message includes the second identifier and the first indication information.

[0431] Optionally, the first NWDAF also sends the first mapping relationship and / or the second mapping relationship to the first AnLF.

[0432] S1108: The first AnLF sends the second identifier, the first indication information, and the information of the first NWDAF to the service consumption network element.

[0433] Optionally, the first AnLF also sends the first mapping relationship and / or the second mapping relationship to the first NWDAF.

[0434] S1109 to S1115 may refer to S1009 to S1015 and will not be elaborated here.

[0435] By using the above method, the first AnLF can discover the NWDAF that supports MTLF and AnLF and has the ability of vertical federated learning, and instruct the NWDAF to initiate a vertical federated task, so that the service consumption network element provides the service corresponding to the analysis identifier through the vertical federated learning model obtained from the vertical federated task.

[0436] It can be understood that, in order to implement the functions in the above embodiments, the first network element and the second network element include the corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should easily realize that, in combination with the units and method steps of each example described in the embodiments disclosed in the present application, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application scenario and design constraint conditions of the technical solution.

[0437] Figure 12 and Figure 13 FIG. is a schematic structural diagram of a possible communication device provided by an embodiment of the present application. These communication devices can be used to implement the functions of the first network element and the second network element in the above method embodiments, and thus can also achieve the beneficial effects possessed by the above method embodiments.

[0438] As Figure 12 shown, the communication device 1200 includes a processing unit 1210 and a transceiver unit 1220. The communication device 1200 is used to implement the functions of the first network element or the second network element in the above method embodiments.

[0439] When the communication device 1200 is used to implement the function of the first network element in the above method embodiment:

[0440] The transceiver unit 1220 is used to send and receive messages; the processing unit 1210 is used to send a first message to a second network element through the transceiver unit 1220 and receive a second message from the second network element; the first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier; the second message includes a first identifier, information of M network elements, and information of a first model; wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model.

[0441] In a possible design, the processing unit 1210 is used to determine at least one network element according to the information of the M network elements; the transceiver unit 1220 is used to send a third message to the at least one network element, and the third message includes the first identifier, and the third message is used to request the vertical federated learning model corresponding to the first identifier to provide services.

[0442] In a possible design, the processing unit 1210 is used to obtain a first intermediate result according to the first model; the transceiver unit 1220 is used to receive at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; the processing unit 1210 is used to determine an output according to the first intermediate result and the at least one intermediate result; the transceiver unit 1220 is used to send the output to a service-consuming network element.

[0443] In a possible design, the processing unit 1210 is used to obtain a first intermediate result according to the first model; the transceiver unit 1220 is used to send the first intermediate result to a third network element, receive an output from the third network element, and send the output to a service-consuming network element; the third network element is one of the at least one network element.

[0444] In a possible design, the transceiver unit 1220 is used to receive a fourth message from the service-consuming network element before sending the first message to the second network element, and the fourth message includes the analysis identifier and information of a first object, wherein the first object corresponds to the analysis identifier; the fourth message is used to request the service corresponding to the analysis identifier.

[0445] In a possible design, the third message further includes the information of the first object.

[0446] In a possible design, the second message further indicates or includes a first mapping relationship, where the first mapping relationship is the mapping relationship between the first object and the object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0447] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0448] In a possible design, the third message further includes the identifier of the first object in the vertical federated task corresponding to the first identifier. The second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. The processing unit 1210 is configured to, if the information of the first object includes the network entity identifier of the first object, determine the identifier of the first object in the vertical federated task corresponding to the first identifier according to the network entity identifier of the first object and the first mapping relationship; if the information of the first object includes the user entity identifier of the first object, determine the identifier of the first object in the vertical federated task corresponding to the first identifier according to the user entity identifier of the first object and the second mapping relationship.

[0449] In a possible design, the second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; or the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0450] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0451] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier; wherein, the j-th network element is the user of the j-th vertical federated learning model, and at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0452] When the communication device 1200 is used to implement the functions of the first network element in the above method embodiments:

[0453] The transceiver unit 1220 is configured to send and receive messages; the processing unit 1210 is configured to receive a first message from the first network element through the transceiver unit 1220 and send a second message to the first network element; the first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier; the second message includes a first identifier, information of M network elements, and information of a first model; wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model.

[0454] In a possible design, the processing unit 1210 is configured to determine at least one of the first identifier, the information of the M network elements, and the information of the first model according to the analysis identifier before sending the second message to the first network element.

[0455] In a possible design, the second message further indicates or includes a first mapping relationship, and the first mapping relationship is a mapping relationship between a first object and an object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0456] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0457] In a possible design, the second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier.

[0458] In a possible design, the second message further includes first indication information. The first indication information indicates that the first model is a vertical federated learning model; or, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or, the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

[0459] In a possible design, the information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

[0460] In a possible design, the j-th network element is any one of the M network elements, and the role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier; wherein, the j-th network element is the user of the j-th vertical federated learning model, and at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model, j is less than or equal to M, and j is a positive integer.

[0461] When the communication device 1200 is used to implement the function of the second network element in the above method embodiment:

[0462] The transceiver unit 1220 is configured to send and receive messages; the processing unit 1210 is configured to send a first message to a second network element via the transceiver unit 1220, receive a second message from the second network element, and the second network element sends a third message; the first message includes an analysis identifier, and the first message is used to request a model for providing services corresponding to the analysis identifier; the second message includes a second identifier and first indication information, the first indication information indicates that the model for providing services corresponding to the analysis identifier is a vertical federated learning model, and the second identifier is used to identify the first message; the third message is used to request services corresponding to the analysis identifier, and the third message includes the second identifier.

[0463] In a possible design, the transceiver unit 1220 is configured to receive an output from the second network element; and send the output to a service-consuming network element.

[0464] In a possible design, the transceiver unit 1220 is configured to receive a fourth message from the service-consuming network element before sending the first message to the second network element, the fourth message includes the analysis identifier and information of a first object, wherein the first object corresponds to the analysis identifier; the fourth message is used to request services corresponding to the analysis identifier.

[0465] In a possible design, the third message further includes the information of the first object.

[0466] In a possible design, the second message further indicates or includes a first mapping relationship, and the first mapping relationship is a mapping relationship between the first object and the object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0467] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship, the first mapping relationship is the identifier of the network entity in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity in the vertical federated task corresponding to the first identifier; it can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0468] In a possible design, the third message further includes the identifier of the first object in the vertical federated learning model; the second message further indicates a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is the identifier of the network entity identifier in the vertical federated learning model, and the second mapping relationship is the identifier of the user entity identifier in the vertical federated learning model; the processing unit 1210 is configured to, if the information of the first object includes the network entity identifier of the first object, determine the identifier of the first object in the vertical federated learning model according to the network entity identifier of the first object and the first mapping relationship; if the information of the first object includes the user entity identifier of the first object, determine the identifier of the first object in the vertical federated learning model according to the user entity identifier of the first object and the second mapping relationship.

[0469] When the communication device 1200 is used to implement the function of the second network element in the above method embodiment:

[0470] The transceiver unit 1220 is configured to send and receive messages. The processing unit 1210 is configured to receive a first message from the first network element through the transceiver unit 1220, send a second message to the first network element, and receive a third message from the first network element; the first message includes an analysis identifier, and the first message is used to request a model for providing a service corresponding to the analysis identifier; the second message includes a second identifier and a first indication information, the first indication information indicates that the model for providing the service corresponding to the analysis identifier is a vertical federated learning model, and the second identifier is used to identify the first message; the third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

[0471] In a possible design, the processing unit 1210 is configured to, before sending the second message to the first network element, the second network element determines the information of the first model according to the analysis identifier; save the corresponding relationship between the second identifier and the information of the first model.

[0472] In a possible design, when the processing unit 1210 determines the information of the first model according to the analysis identifier, the second network element determines the information of the first model, the information of M network elements, and the first identifier according to the analysis identifier; when the second network element saves the correspondence between the second identifier and the information of the first model, it saves the correspondence between the second identifier and the information of the first model, the information of the M network elements, and the first identifier; wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models correspond to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model.

[0473] In a possible design, after receiving the third message from the first network element, the processing unit 1210 determines the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; determines at least one network element according to the information of the M network elements; the transceiver unit 1220 is configured to send a fifth message to the at least one network element, the fifth message includes the first identifier, and the fifth message is used to request a vertical federated learning model corresponding to the first identifier to provide services; the processing unit 1210 is configured to obtain a first intermediate result according to the first model; the transceiver unit 1220 is configured to receive at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; the processing unit 1210 is configured to determine the output according to the first intermediate result and the at least one intermediate result.

[0474] In a possible design, after receiving the third message from the first network element, the processing unit 1210 determines the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; determines at least one network element according to the information of the M network elements; the transceiver unit 1220 is configured to send a fifth message to the at least one network element, the fifth message includes the first identifier, and the third message is used to request a vertical federated learning model corresponding to the first identifier to provide services; the processing unit 1210 is configured to obtain a first intermediate result according to the first model; the transceiver unit 1220 is configured to send the first intermediate result to the third network element; the second network element receives the output from the third network element.

[0475] In a possible design, the transceiver unit 1220 is configured to send the output to the first network element.

[0476] In a possible design, the third message further includes information of a first object, and / or an identifier of the first object in the vertical federated task corresponding to the first identifier.

[0477] In a possible design, the second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is an identifier of a network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity identifier in the vertical federated task corresponding to the first identifier.

[0478] In a possible design, the second message further indicates or includes a first mapping relationship. The first mapping relationship is a mapping relationship between a first object and an object of the first object in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is a mapping relationship between the identifier of the first object and the identifier of the first object in the vertical federated task corresponding to the first identifier. The first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the identifier of the first object and the first mapping relationship.

[0479] In a possible design, the second message further indicates or includes a first mapping relationship and / or a second mapping relationship. The first mapping relationship is an identifier of a network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity identifier in the vertical federated task corresponding to the first identifier. It can be understood that the first mapping relationship is the relationship between the network entity identifier and the identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the relationship between the user entity identifier and the identifier in the vertical federated task corresponding to the first identifier.

[0480] For a more detailed description of the above processing unit 1210 and transceiver unit 1220, reference can be directly made to the relevant descriptions in the above method embodiments and will not be elaborated here.

[0481] As Figure 13 shown, the communication device 1300 includes a processor 1310 and an interface circuit 1320. The processor 1310 and the interface circuit 1320 are coupled to each other. It can be understood that the interface circuit 1320 can be a transceiver or an input / output interface. Optionally, the communication device 1300 may further include a memory 1330 for storing instructions executed by the processor 1310 or storing input data required for the processor 1310 to run instructions or storing data generated after the processor 1310 runs instructions.

[0482] When the communication device 1300 is used to implement Figure 5When implementing the method shown, the processor 1310 is used to implement the functions of the above-mentioned processing unit 1210, and the interface circuit 1320 is used to implement the functions of the above-mentioned transceiver unit 1220.

[0483] It can be understood that the processor in the embodiments of the present application may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0484] In the present application, another example of a device is provided. The notification device includes at least one processor and at least one memory. The at least one processor is coupled to the at least one memory. The at least one memory is used to store instructions. When the instructions are executed by the at least one processor, the communication device executes the method in the above embodiments. Taking the communication device including one processor and one memory as an example, as Figure 13 shown, the communication device 1300 includes a processor 1310 and a memory 1330. The processor 1310 is coupled to the memory 1330. Instructions are stored in the memory 1330. When the instructions stored in the memory 1330 are executed by the processor 1310, the communication device 1300 executes the methods executed by each network element in the above embodiments.

[0485] The method steps in the embodiments of the present application can be implemented in hardware or in software instructions executable by a processor. The software instructions may be composed of corresponding software modules. The software modules may be stored in a random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, register, hard disk, removable hard disk, CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. The storage medium may also be a part of the processor. The processor and the storage medium may be located in the ASIC. Additionally, the ASIC may be located in the above-mentioned network element. The processor and the storage medium may also exist as discrete components in the above-mentioned network element.

[0486] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in the form of a computer program product in whole or in part. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, or other programmable devices. The computer program or instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless manner. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it can also be an optical medium, such as a digital video disc; or it can be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or can include both volatile and non-volatile types of storage media.

[0487] In this application, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. In the textual description of this application, the character " / " generally represents an "or" relationship between the associated objects before and after; in the formulas of this application, the character " / " represents a "division" relationship between the associated objects before and after. "Including at least one of A, B, and C" can represent: including A; including B; including C; including A and B; including A and C; including B and C; including A, B, and C.

[0488] It can be understood that the various numerical numbers involved in the embodiments of this application are only for the convenience of description and are not used to limit the scope of the embodiments of this application. The magnitudes of the serial numbers of the above processes do not mean the order of execution, and the execution order of each process should be determined by its function and internal logic.

Claims

1. A communication method, characterized in that, The method includes: A first network element sends a first message to a second network element; the first message includes an analysis identifier, and the first message is used to request a model for providing the service corresponding to the analysis identifier; The first network element receives a second message from the second network element, and the second message includes a first identifier, information of M network elements, and information of a first model; Wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model.

2. The method according to claim 1, characterized in that, It further includes: The first network element determines at least one network element according to the information of the M network elements; The first network element sends a third message to the at least one network element, and the third message includes the first identifier, and the third message is used to request the vertical federated learning model corresponding to the first identifier to provide services.

3. The method according to claim 2, wherein The method further includes: The first network element obtains a first intermediate result according to the first model; The first network element receives at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; The first network element determines an output according to the first intermediate result and the at least one intermediate result; The first network element sends the output to a service-consuming network element.

4. The method according to claim 2, wherein The method further includes: The first network element obtains a first intermediate result according to the first model; The first network element sends the first intermediate result to a third network element; the third network element is one of the at least one network element; The first network element receives an output from the third network element; The first network element sends the output to a service-consuming network element.

5. The method according to any one of claims 2-4, characterized in that, Before the first network element sends the first message to the second network element, the method further includes: The first network element receives a fourth message from the service-consuming network element, and the fourth message includes the analysis identifier and information of a first object, wherein the first object corresponds to the analysis identifier; the fourth message is used to request the service corresponding to the analysis identifier.

6. The method according to claim 5, wherein The third message further includes the information of the first object.

7. The method according to claim 5 or 6, characterized in that, The third message further includes an identifier of the first object in the vertical federated task corresponding to the first identifier; the second message further includes a first mapping relationship and / or a second mapping relationship, the first mapping relationship is an identifier of a network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity identifier in the vertical federated task corresponding to the first identifier; The method further includes: If the information of the first object includes the network entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the network entity identifier of the first object and the first mapping relationship; Alternatively, if the information of the first object includes the user entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated task corresponding to the first identifier according to the user entity identifier of the first object and the second mapping relationship.

8. The method according to any one of claims 1-7, characterized in that, The second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; alternatively, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

9. The method according to any one of claims 1-8, characterized in that, The information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

10. The method according to claim 9, wherein The j-th network element is any one of the M network elements. The at least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model. The j-th network element uses the j-th vertical federated learning model to provide services. The role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier, where j is less than or equal to M and j is a positive integer.

11. A communication method, characterized in that, The method includes: The second network element receives a first message from the first network element; the first message includes an analysis identifier, and the first message is used to request a model for providing services corresponding to the analysis identifier. The second network element sends a second message to the first network element, and the second message includes a first identifier, information of M network elements, and information of a first model. Wherein, the first identifier is associated with the analysis identifier, and there are at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model.

12. The method according to claim 11, wherein Before the second network element sends the second message to the first network element, it further includes: The second network element determines at least one of the first identifier, the information of the M network elements, and the information of the first model according to the analysis identifier.

13. The method according to claim 11 or 12, characterized in that, The second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of the network entity identifier in the vertical federated task corresponding to the first identifier, and the second mapping relationship is the identifier of the user entity identifier in the vertical federated task corresponding to the first identifier.

14. The method according to any one of claims 11-13, characterized in that The second message further includes first indication information, where the first indication information indicates that the first model is a vertical federated learning model; alternatively, the first indication information indicates that the model corresponding to the first identifier is a vertical federated learning model; or the first indication information indicates that the task corresponding to the first identifier is a vertical federated task.

15. The method according to any one of claims 11-14, characterized in that, The information of the M network elements includes at least one of the network entity identifiers of the M network elements, the subscription association addresses of the M network elements, and the role information of the M network elements.

16. The method according to claim 15, wherein The j-th network element is any one of the M network elements. At least two vertical federated learning models corresponding to the first identifier include the j-th vertical federated learning model. The j-th network element uses the j-th vertical federated learning model to provide services. The role information of the j-th network element indicates the role of the producer of the j-th vertical federated learning model in the vertical federated task corresponding to the first identifier. j is less than or equal to M and j is a positive integer.

17. A communication method, characterized in that, The method includes: The first network element sends a first message to the second network element; the first message includes an analysis identifier, and the first message is used to request a model that provides the service corresponding to the analysis identifier; The first network element receives a second message from the second network element. The second message includes a second identifier and first indication information. The first indication information indicates that the model that provides the service corresponding to the analysis identifier is a vertical federated learning model. The second identifier is used to identify the first message; The first network element sends a third message to the second network element. The third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

18. The method according to claim 17, wherein The method further includes: The first network element receives an output from the second network element; The first network element sends the output to a service consumption network element.

19. The method according to claim 17 or 18, characterized in that, Before the first network element sends the first message to the second network element, the method further includes: The first network element receives a fourth message from the service consumption network element. The fourth message includes the analysis identifier and information of a first object, where the first object corresponds to the analysis identifier; the fourth message is used to request the service corresponding to the analysis identifier.

20. The method according to claim 19, characterized in that, The third message further includes the information of the first object.

21. The method according to claim 19 or 20, characterized in that The third message further includes an identifier of the first object in the vertical federated learning model; the second message further indicates a first mapping relationship and / or a second mapping relationship. The first mapping relationship is the identifier of a network entity identifier in the vertical federated learning model, and the second mapping relationship is the identifier of a user entity identifier in the vertical federated learning model; The method further includes: If the information of the first object includes the network entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated learning model according to the network entity identifier of the first object and the first mapping relationship; Alternatively, if the information of the first object includes the user entity identifier of the first object, the first network element determines the identifier of the first object in the vertical federated learning model according to the user entity identifier of the first object and the second mapping relationship.

22. A communication method, characterized in that The method includes: The second network element receives a first message from the first network element; the first message includes an analysis identifier, and the first message is used to request a model that provides the service corresponding to the analysis identifier; The second network element sends a second message to the first network element. The second message includes a second identifier and first indication information. The first indication information indicates that the model that provides the service corresponding to the analysis identifier is a vertical federated learning model. The second identifier is used to identify the first message; The second network element receives a third message from the first network element, where the third message is used to request the service corresponding to the analysis identifier, and the third message includes the second identifier.

23. The method according to claim 22, wherein Before the second network element sends a second message to the first network element, it further includes: The second network element determines information of a first model according to the analysis identifier; The second network element stores the corresponding relationship between the second identifier and the information of the first model.

24. The method according to claim 23, wherein, The second network element determines information of a first model according to the analysis identifier, including: The second network element determines the information of the first model, the information of M network elements, and a first identifier according to the analysis identifier; The second network element stores the corresponding relationship between the second identifier and the information of the first model, including: The second network element stores the corresponding relationship between the second identifier and the information of the first model, the information of the M network elements, and the first identifier; Wherein, the first identifier is associated with the analysis identifier, and at least two vertical federated learning models corresponding to the first identifier; the M network elements use the at least two vertical federated learning models to provide services, M is a positive integer, and the at least two vertical federated learning models include the first model.

25. The method according to claim 24, wherein After the second network element receives the third message from the first network element, it further includes: The second network element determines the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; The second network element determines at least one network element according to the information of the M network elements; The second network element sends a fifth message to the at least one network element, where the fifth message includes the first identifier, and the fifth message is used to request the vertical federated learning model corresponding to the first identifier to provide services; The second network element obtains a first intermediate result according to the first model; The second network element receives at least one intermediate result from the at least one network element, and the at least one network element corresponds to the at least one intermediate result one by one; The second network element determines the output according to the first intermediate result and the at least one intermediate result.

26. The method according to claim 24, wherein After the second network element receives the third message from the first network element, it further includes: The second network element determines the information of the first model, the information of the M network elements, and the first identifier according to the second identifier; The second network element determines at least one network element according to the information of the M network elements; The second network element sends a fifth message to the at least one network element, where the fifth message includes the first identifier, and the third message is used to request the vertical federated learning model corresponding to the first identifier to provide services; The second network element obtains a first intermediate result according to the first model; The second network element sends the first intermediate result to the third network element; The second network element receives the output from the third network element.

27. The method according to claim 25 or 26, characterized in that, It further includes: The second network element sends the output to the first network element.

28. The method according to any one of claims 24-27, characterized in that, The third message further includes information of a first object, and / or an identifier of the first object in the vertical federated task corresponding to the first identifier.

29. The method according to claim 28, wherein, The second message further indicates a first mapping relationship and / or a second mapping relationship, where the first mapping relationship is an identifier of a network entity in a vertical federated task corresponding to the first identifier, and the second mapping relationship is an identifier of a user entity in a vertical federated task corresponding to the first identifier.

30. A communication device, characterized in that, It includes units or modules for performing the method according to any one of claims 1 to 29.

31. A communication device, characterized in that, The communication device includes at least one processor; the at least one processor is configured to perform the method according to any one of claims 1 to 29.

32. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a program that, when running on a device, causes the device to perform the method according to any one of claims 1 to 29.

33. A computer program product, characterized in that, The computer program product includes a program or instructions that, when executed by a device, cause the device to perform the method according to any one of claims 1 to 29.

Images