Model inference method, server, NEF, client and computer program product

Abstract

The present disclosure relates to the technical field of wireless communications. Disclosed are a model inference method, a server, an NEF, a client and a computer program product. The model inference method is executed by a server, and comprises: determining for each client among a plurality of clients information related to vertical federated learning (VFL) inference, wherein the information which is related to VFL inference and corresponds to each client comprises at least one of a VFL association identifier, first sample indication information and first feature indication information which correspond to the client; and sending first request information to each client, so that the client performs inference, wherein the first request information comprises the information which is related to VFL inference and corresponds to the client.

Description

Inference method of model, server, NEF, client and computer program product

[0001] Cross-reference to Related Applications

[0002] This application is based on the application with CN application number 202411096529.8, filed on August 9, 2024, and claims priority thereto, the disclosure of which is hereby incorporated by reference in its entirety. TECHNICAL FIELD

[0003] The present disclosure relates to the technical field of wireless communication, in particular to an inference method of model, a server, a NEF, a client, a computer readable storage medium and a computer program product. BACKGROUND

[0004] With the development of federated learning technology, the concept of vertical federated learning is introduced, which is different from the existing horizontal federated learning technology.

[0005] Vertical federated learning is divided into training and inference processes, and the inference process occurs after model training. Due to data isolation, vertical federated learning has the same data samples and different sample characteristics, that is, data samples of different characteristics are distributed in different network elements, and these original data cannot be exchanged in the inference process, so compared with the traditional inference process, the inference process of vertical federated learning needs the participation of all participants, that is, distributed inference. SUMMARY

[0006] According to some embodiments of the first aspect of the present disclosure, an inference method of model is provided, executed by a server, comprising: determining, for each client in a plurality of clients, related information of vertical federated learning (VFL) inference, wherein the related information of VFL inference corresponding to each client comprises at least one of a VFL association identifier corresponding to each client, first sample indication information and first feature indication information; sending first request information to each client so that each client performs inference, wherein the first request information comprises the related information of VFL inference corresponding to each client.

[0007] In some embodiments, the VFL association identifier is used to associate the process of VFL and / or the VFL model; the first sample indication information is used to indicate the data sample; and / or the first feature indication information is used to indicate the data feature.

[0008] In some embodiments, the information related to the VFL inference corresponding to each client further includes at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator, the analysis identifier is used to identify a requested analysis or identify that a requested analysis and / or an analysis result is generated using a VFL manner, the VFL indicator is used to indicate a model trained by VFL, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and data network access identification.

[0009] In some embodiments, determining the information related to the vertical federated learning (VFL) inference for each of the plurality of clients includes:

[0010] According to a preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients is divided to determine the information related to the VFL inference for each client, wherein the preset rule includes at least one of an internal configuration of the server, an internal logic of the server, a network element type of each client, related information of each client obtained from other network elements in a process of finding the client, and information associated by each client in a training process.

[0011] In some embodiments, according to the preset rule, dividing at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients to determine the information related to the VFL inference for each client includes at least one of: dividing a set of user equipment identifiers in the sample indication information corresponding to the plurality of clients according to the preset rule to determine, for each client, a set of user equipment identifiers corresponding to the client in the first sample indication information; dividing sample identifiers in the sample indication information corresponding to the plurality of clients according to the preset rule to determine, for each client, sample identifiers corresponding to the client in the first sample indication information; dividing feature identifiers in the feature indication information corresponding to the plurality of clients according to the preset rule to determine, for each client, feature identifiers corresponding to the client in the first feature indication information; dividing quality of experience indicator information in the feature indication information corresponding to the plurality of clients according to the preset rule to determine, for each client, quality of experience indicator information corresponding to the client in the first feature indication information; dividing a region of interest in the filter information corresponding to the plurality of clients according to the preset rule to determine, for each client, a region of interest corresponding to the client in the filter information; and dividing quality of experience indicator information in the filter information corresponding to the plurality of clients according to the preset rule to determine, for each client, quality of experience indicator information corresponding to the client in the filter information.

[0012] In some embodiments, the inference method further comprises: performing local VFL inference according to data in the server to obtain an intermediate result of the local VFL inference.

[0013] In some embodiments, the inference method further comprises: receiving first response information sent by each client, wherein the first response information corresponding to each client comprises at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier; and generating an inference result according to the first response information sent by each client and / or the intermediate result of the local VFL inference.

[0014] In some embodiments, the inference method further comprises: receiving first response information sent by a network exposure function (NEF), wherein the first response information is obtained by the NEF aggregating at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier; and generating an inference result according to the first response information and / or the intermediate result of the local VFL inference.

[0015] In some embodiments, the inference method further comprises: receiving second request information sent by a consuming network function (NF), wherein the second request information is used to request an analysis result, the second request information comprises at least one of an analysis identifier and a VFL indicator, the analysis identifier is used to identify a requested analysis and / or an analysis result generated using a VFL manner, and the VFL indicator is used to indicate a VFL trained model, and wherein sending the first request information to each client comprises: sending the first request information to each client according to the second request information.

[0016] In some embodiments, the second request information further comprises filter information, wherein the filter information comprises at least one of a single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0017] In some embodiments, the inference method further comprises: sending second response information to the consuming NF, wherein the second response information comprises at least one of an analysis result and second validity time information, the analysis result is generated according to the inference result, and the second validity time information is used to indicate a validity time of the inference result.

[0018] In some embodiments, the first response information further comprises first validity time information used to indicate a validity time of the intermediate result of the VFL inference.

[0019] In some embodiments, the first sample indication information is used for indicating the data sample; the second sample indication information is used for indicating the data sample; the first sample indication information comprises at least one of a user equipment identity, a generic public subscriber identity (GPSI), a subscription permanent identifier (SUPI), location information, a sample identity, and a group identity, and / or the second sample indication information comprises at least one of a user equipment identity, a generic public subscriber identity (GPSI), a subscription permanent identifier (SUPI), location information, a sample identity, and a group identity.

[0020] In some embodiments, the first feature indication information is used for indicating a data feature used by the server and / or a data feature used by the client; the second feature indication information is used for indicating a data feature used by the server and / or a data feature used by the client; the first feature indication information comprises at least one of service experience information, quality of experience (QoE) information, voice quality information, and a feature identity, and / or the second feature indication information comprises at least one of service experience information, quality of experience (QoE) information, voice quality information, and a feature identity.

[0021] In some embodiments, the reasoning method further comprises: collecting the data sample from the corresponding network element according to the first request information.

[0022] In some embodiments, the analysis result is in the form of an analysis ID output, comprises at least one of various data of the analysis ID output, or comprises at least one of quality of experience (QoE) information, latency information, and 5G quality of service (QoS) indicators.

[0023] In some embodiments, the consuming NF is an analysis logic function (AnLF), and the server is an application function (AF).

[0024] In some embodiments, the server is a network data analytics function (NWDAF), and each client is an application function (AF), or the server and each client are different NWDAFs, or the server is an AF, and each client is an NWDAF.

[0025] In some embodiments, in the case of the AF being a non-trusted AF, sending the first request information to each client comprises: sending the first request information to each client through a network exposure function (NEF), wherein, in the case of the service end being a non-trusted AF, the VFL association identifier in the first request information is an external VFL association identifier, the first sample indication information comprises first external sample indication information, and / or the first feature indication information comprises first external feature indication information, the NEF maps the external VFL association identifier in the first request information to an internal VFL association identifier, maps the first external sample indication information to first internal sample indication information, and / or maps the first external feature indication information to first internal feature indication information, in the case of there being a non-trusted AF among the plurality of clients, for each client that is a non-trusted AF, the VFL association identifier in the first request information is an internal VFL association identifier, the first sample indication information comprises first internal sample indication information, and / or the first feature indication information comprises first internal feature indication information, the NEF maps the first internal VFL association identifier in the first request information to a first external VFL association identifier, maps the first internal sample indication information to first external sample indication information, and / or maps the first internal feature indication information to first external feature indication information.

[0026] In some embodiments, the inference method further comprises: receiving first response information sent by each client forwarded through a network exposure function (NEF), wherein the first response information corresponding to each client comprises at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier; and generating an inference result according to the first response information sent by each client and / or an intermediate result of local VFL inference.

[0027] In some embodiments, the inference method further comprises: in the case that there are untrusted AFS in the plurality of clients, for each client of the untrusted AFS, the intermediate result of the VFL inference in the first response information is an external intermediate result, the VFL association identifier is an external VFL association identifier, the second sample indication information comprises second external sample indication information, and / or the second feature indication information comprises second external feature indication information, the NEF maps the external intermediate result in the first response information to an internal intermediate result, maps the external VFL association identifier to an internal VFL association identifier, maps the second external sample indication information to second internal sample indication information, and / or maps the second external feature indication information to second internal feature indication information; in the case that the service end is an untrusted AF, the intermediate result of the VFL inference in the first response information is an internal intermediate result, the VFL association identifier is an internal VFL association identifier, the second sample indication information comprises second internal sample indication information, and / or the second feature indication information comprises second internal feature indication information, the NEF maps the internal intermediate result in the first response information to an external intermediate result, maps the internal VFL association identifier to an external VFL association identifier, maps the second internal sample indication information to second external sample indication information, and / or maps the second internal feature indication information to second external feature indication information.

[0028] In some embodiments, generating the inference result according to the first response information sent by each client and / or the intermediate result of the local VFL inference comprises: aggregating the first response information sent by each client and the intermediate result of the local VFL inference to generate the inference result.

[0029] According to some embodiments of the second aspect of the present disclosure, an inference method of a model is provided, executed by an NEF, comprising: receiving third request information sent by a service end, wherein the third request information comprises a VFL association identifier and at least one of sample indication information and feature indication information corresponding to a plurality of clients; determining, according to the third request information, relevant information of vertical federated learning VFL inference for each client in the plurality of clients, wherein the relevant information of the VFL inference corresponding to each client comprises at least one of a VFL association identifier corresponding to each client, third sample indication information, and third feature indication information; and sending third sub-request information comprising the relevant information of the VFL inference corresponding to each client to each client so as to perform inference by each client.

[0030] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model; the sample indication information is used to indicate the data sample; and / or the feature indication information is used to indicate the data feature.

[0031] In some embodiments, the information related to the VFL inference corresponding to each client further includes at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator, the analysis identifier is used to identify a requested analysis or to identify that the requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a model trained in a VFL manner, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and data network access identification.

[0032] In some embodiments, determining, according to the third request information, the information related to the vertical federated learning (VFL) inference for each of the plurality of clients includes: dividing, according to a preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients, and determining the information related to the VFL inference for each of the plurality of clients, wherein the preset rule includes at least one of an internal configuration of the server, an internal logic of the server, a network element type of each of the plurality of clients, information related to each of the plurality of clients obtained from other network elements in a process of finding the plurality of clients, and information associated with each of the plurality of clients in a training process.

[0033] In some embodiments, dividing, according to the preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients, and determining the information related to the VFL inference for each of the plurality of clients includes at least one of: dividing, according to the preset rule, a set of user equipment identifiers in the sample indication information corresponding to the plurality of clients, and determining, for each of the plurality of clients, a set of user equipment identifiers corresponding to the client in the third sample indication information; dividing, according to the preset rule, sample identifiers in the sample indication information corresponding to the plurality of clients, and determining, for each of the plurality of clients, sample identifiers corresponding to the client in the third sample indication information; dividing, according to the preset rule, feature identifiers in the feature indication information corresponding to the plurality of clients, and determining, for each of the plurality of clients, feature identifiers corresponding to the client in the third feature indication information; dividing, according to the preset rule, quality of experience indicator information in the feature indication information corresponding to the plurality of clients, and determining, for each of the plurality of clients, quality of experience indicator information corresponding to the client in the third feature indication information; dividing, according to the preset rule, a region of interest in the filter information corresponding to the plurality of clients, and determining, for each of the plurality of clients, a region of interest corresponding to the client in the filter information; and dividing, according to the preset rule, quality of experience indicator information in the filter information corresponding to the plurality of clients, and determining, for each of the plurality of clients, quality of experience indicator information corresponding to the client in the filter information.

[0034] In some embodiments, the inference method further comprises: receiving third sub-response information sent by each client, wherein the third sub-response information comprises at least one of an intermediate result of VFL inference of each client, fourth feature indication information, fourth sample indication information, and a VFL association identifier; in the case that the NEF has aggregation capability, aggregating the third sub-response information to obtain third response information; and sending the third response information to the server, so that the server generates an inference result according to the third response information and / or the intermediate result of local VFL inference.

[0035] In some embodiments, the third response information further comprises third validity time information for indicating a validity time of the intermediate result of VFL inference.

[0036] In some embodiments, the third sample indication information is used for indicating a data sample, the fourth sample indication information is used for indicating a data sample, the third sample indication information comprises at least one of a user equipment identifier, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier, and / or the fourth sample indication information comprises at least one of a user equipment identifier, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0037] In some embodiments, the third feature indication information is used for indicating a data feature used by the server and / or a data feature used by the client, the fourth feature indication information is used for indicating a data feature used by the server and / or a data feature used by the client, the third feature indication information comprises at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier, and / or the fourth feature indication information comprises at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

[0038] In some embodiments, the server is a network data analytics function NWDAF, each client is an application function AF, or the server and each client are different NWDAFs, or the server is an AF and each client is an NWDAF.

[0039] In some embodiments, the reasoning method further comprises: in the case that the non-trusted AF exists in the plurality of clients, for each client that is a non-trusted AF, the VFL correlation identifier in the third sub-request information is an internal VFL correlation identifier, the third sample indication information comprises third internal sample indication information, and / or the third feature indication information comprises third internal feature indication information, mapping the internal VFL correlation identifier in the third sub-request information to an external VFL correlation identifier, mapping the third internal sample indication information to third external sample indication information, and / or mapping the third internal feature indication information to third external feature indication information.

[0040] In some embodiments, the reasoning method further comprises: in the case that the NEF does not have aggregation capability, sending the third sub-response information to the server, so that the server generates the reasoning result according to the third sub-response information and / or the intermediate result of the local VFL reasoning.

[0041] In some embodiments, wherein, in the case that the non-trusted AF exists in the plurality of clients, for each client that is a non-trusted AF, the intermediate result of the VFL reasoning in the third sub-response information is an external intermediate result, the VFL correlation identifier is an external VFL correlation identifier, the fourth sample indication information comprises fourth external sample indication information, and / or the fourth feature indication information comprises fourth external feature indication information, mapping the external intermediate result in the third sub-response information to an internal intermediate result, mapping the external VFL correlation identifier to an internal VFL correlation identifier, mapping the fourth external sample indication information to fourth internal sample indication information, and / or mapping the fourth external feature indication information to fourth internal feature indication information, in the case that the server is a non-trusted AF, the intermediate result of the VFL reasoning in the third sub-response information is an internal intermediate result, the VFL correlation identifier is an internal VFL correlation identifier, the fourth sample indication information comprises fourth internal sample indication information, and / or the fourth feature indication information comprises fourth internal feature indication information, mapping the internal intermediate result in the third sub-response information to an external intermediate result, mapping the internal VFL correlation identifier to an external VFL correlation identifier, mapping the fourth internal sample indication information to fourth external sample indication information, and / or mapping the fourth internal feature indication information to fourth external feature indication information.

[0042] According to some embodiments of the third aspect of the present disclosure, a model inference method is provided, executed by a client, comprising: receiving first request information sent by a server, wherein the first request information comprises VFL inference related information corresponding to the client, the VFL inference related information corresponding to the client comprises at least one of a VFL association identifier corresponding to the client, first sample indication information, and first feature indication information, and the VFL inference related information corresponding to the client is divided for the client by the server from a plurality of VFL inference related information corresponding to a plurality of clients; and performing VFL inference according to the first request information.

[0043] In some embodiments, the VFL association identifier is used to associate a process of the VFL and / or a VFL model; the first sample indication information is used to indicate a data sample; and / or the first feature indication information is used to indicate a data feature.

[0044] In some embodiments, the inference method further comprises: sending first response information to the server, wherein the first response information comprises at least one of an intermediate result of the VFL inference corresponding to the client, second sample indication information, second feature indication information, and the VFL association identifier.

[0045] In some embodiments, the VFL inference related information corresponding to the client further comprises at least one of an interoperation indicator corresponding to the client, an analysis identifier, filter information, and a VFL indicator, the analysis identifier is used to identify a requested analysis or to identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a VFL trained model, and the filter information comprises at least one of single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0046] In some embodiments, the first response information further comprises first validity time information used to indicate a validity time of the intermediate result of the VFL inference.

[0047] In some embodiments, the first sample indication information is used to indicate a data sample; the second sample indication information is used to indicate a data sample; the first sample indication information comprises at least one of a user equipment identifier, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier, and / or the second sample indication information comprises at least one of a user equipment identifier, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0048] In some embodiments, the first feature indication information is used to indicate the data features used by the server and / or the data features used by the client, and the second feature indication information is used to indicate the data features used by the server and / or the data features used by the client; the first feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier, and / or the second feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier.

[0049] In some embodiments, the inference method further comprises: collecting data samples from the corresponding network element according to the first request information.

[0050] In some embodiments, the server is a network data analysis function (NWDAF), and the client is an application function (AF), or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

[0051] According to some embodiments of the fourth aspect of the present disclosure, an inference method of a model is provided, executed by a client, comprising: receiving third sub-request information sent by a network exposure function (NEF), wherein the third sub-request information comprises related information of VFL inference corresponding to the client, the related information of VFL inference corresponding to the client comprises at least one of a VFL association identifier corresponding to the client, third sample indication information, and third feature indication information, and the related information of VFL inference corresponding to the client is divided for the client from a plurality of related information of VFL inference corresponding to clients obtained by the NEF from a server; and performing inference according to the third sub-request information.

[0052] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model, the third sample indication information is used to indicate data samples, and / or the third feature indication information is used to indicate data features.

[0053] In some embodiments, the inference method further comprises: sending third sub-response information to the NEF, wherein the third sub-response information comprises at least one of an intermediate result of VFL inference of the client, fourth feature indication information, fourth sample indication information, and a VFL association identifier, so that the NEF generates response information according to the third sub-response information and sends the response information to the server.

[0054] In some embodiments, the related information of the VFL inference corresponding to the client further includes at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator corresponding to the client, the analysis identifier is used to identify a requested analysis or identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a model trained by VFL, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0055] In some embodiments, the third sample indication information is used to indicate a data sample, and the fourth sample indication information is used to indicate a data sample, the third sample indication information includes at least one of a user equipment identifier, a generic public subscriber identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier, and / or the fourth sample indication information includes at least one of a user equipment identifier, a generic public subscriber identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0056] In some embodiments, the third feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client, and the fourth feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client, the third feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier, and / or the fourth feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

[0057] In some embodiments, the server is a network data analytics function NWDAF, and the client is an application function AF, or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

[0058] According to some embodiments of the fifth aspect of the present disclosure, a server is provided, including: a first determination unit configured to determine, for each client in a plurality of clients, related information of vertical federated learning VFL inference, wherein the related information of the VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, first sample indication information, and first feature indication information; and a first sending unit configured to send, to each client, first request information so as to perform inference by each client, wherein the first request information includes the related information of the VFL inference corresponding to each client.

[0059] According to some embodiments of the sixth aspect of the present disclosure, a NEF is provided, comprising: a first receiving unit configured to receive third request information sent by a server, wherein the third request information comprises at least one of a VFL association identifier, sample indication information corresponding to a plurality of clients, and feature indication information; a second determining unit configured to determine, according to the third request information, information related to VFL inference for each client in the plurality of clients, wherein the information related to VFL inference corresponding to each client comprises at least one of a VFL association identifier corresponding to each client, third sample indication information, and third feature indication information; and a second sending unit configured to send third sub-request information to each client so as to perform inference by each client, wherein the third sub-request information comprises the information related to VFL inference corresponding to each client.

[0060] According to some embodiments of the seventh aspect of the present disclosure, a client is provided, comprising: a second receiving unit configured to receive first request information sent by a server, wherein the first request information comprises information related to VFL inference corresponding to the client, the information related to VFL inference corresponding to the client comprises at least one of a VFL association identifier corresponding to the client, first sample indication information, and first feature indication information, and the information related to VFL inference corresponding to the client is divided for the client by the server from information related to VFL inference corresponding to a plurality of clients; and a first inference unit configured to perform VFL inference according to the first request information.

[0061] According to some embodiments of the eighth aspect of the present disclosure, a client is provided, comprising: a third receiving unit configured to receive third sub-request information sent by a network exposure function (NEF), wherein the third sub-request information comprises information related to VFL inference corresponding to the client, the information related to VFL inference corresponding to the client comprises at least one of a VFL association identifier corresponding to the client, third sample indication information, and third feature indication information, and the information related to VFL inference corresponding to the client is divided for the client by the NEF from information related to VFL inference corresponding to a plurality of clients obtained by the NEF from a server; and a second inference unit configured to perform inference according to the third sub-request information.

[0062] According to some embodiments of the ninth aspect of the present disclosure, an electronic device is provided, comprising: a memory and a processor coupled to the memory, the processor being configured to execute, based on instructions stored in the memory, the inference method of the model in any one of the embodiments of the present disclosure.

[0063] According to some embodiments of the tenth aspect of the present disclosure, a computer readable storage medium is provided, having stored thereon a computer program, which, when executed by a processor, implements the inference method of the model in any one of the embodiments of the present disclosure.

[0064] According to some embodiments of the twelfth aspect of the present disclosure, there is provided a computer program comprising instructions which, when executed by a processor, cause the processor to perform the inference method of the model according to any one of the embodiments of the present disclosure.

[0065] According to some embodiments of the twelfth aspect of the present disclosure, there is provided a computer program comprising instructions which, when executed by a processor, cause the processor to perform the inference method of the model according to any one of the embodiments of the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0066] The accompanying drawings, which constitute a part of this specification, illustrate embodiments of the present disclosure and serve to explain the principles of the present disclosure.

[0067] The present disclosure can be understood more readily by reference to the following detailed description, when taken in connection with the accompanying drawings, and wherein:

[0068] FIG. 1 shows a flowchart of some embodiments of the inference method of the model of the present disclosure.

[0069] FIG. 2 shows a flowchart of some other embodiments of the inference method of the model of the present disclosure.

[0070] FIG. 3 shows a flowchart of some further embodiments of the inference method of the model of the present disclosure.

[0071] FIG. 4 shows a flowchart of some yet other embodiments of the inference method of the model of the present disclosure.

[0072] FIG. 5 shows a schematic diagram of some embodiments of the server of the present disclosure.

[0073] FIG. 6 shows a schematic diagram of some embodiments of the NEF of the present disclosure.

[0074] FIG. 7 shows a schematic diagram of some embodiments of the client of the present disclosure.

[0075] FIG. 8 shows a schematic diagram of some other embodiments of the client of the present disclosure.

[0076] FIG. 9 shows a schematic diagram of some embodiments of the electronic device of the present disclosure.

[0077] FIG. 10 shows a schematic diagram of some embodiments of the inference method of the model of the present disclosure.

[0078] FIG. 11 shows a schematic diagram of some other embodiments of the inference method of the model of the present disclosure.

[0079] FIG. 12 shows a schematic diagram of some further embodiments of the inference method of the model of the present disclosure. DETAILED DESCRIPTION

[0080] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. Note that the relative arrangement of the components and steps illustrated in these embodiments, numerical expressions, and numerical values are not limiting of the scope of the present disclosure unless otherwise specifically stated.

[0081] It should also be understood that the sizes of the regions illustrated in the drawings are chosen primarily for convenience and clarity of presentation and are not necessarily to scale.

[0082] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the scope of the disclosure, its application, or uses.

[0083] Techniques, methods, and devices known to those of ordinary skill in the relevant art can not be discussed in detail herein. However, where appropriate, techniques, methods, and devices should be considered part of the description of the present disclosure.

[0084] In all of the examples shown and discussed herein, any specific values should be interpreted as merely illustrative and not as a limitation on the scope of the disclosure. Thus, other examples of the exemplary embodiments can have different values.

[0085] It should be noted that like reference numerals and letters refer to like items in the following drawings, and thus, once an item is defined in one drawing, it need not be discussed further in subsequent drawings.

[0086] The inventors of the present disclosure found that the above-mentioned related art has the following problem: how to improve the feasibility of distributed inference in the inference process of vertical federated learning.

[0087] Vertical federated learning (VFL) has the same data samples and different sample features. In the 5GC (5 Generation Core, 5G core network), for example, the data in the AF (Application Function) and the NWDAF (Network Data Analytics Function) for the user's quality of experience index information has different data features and the same data samples.

[0088] Therefore, in the inference process of vertical federated learning, it is necessary for each participant to participate in inference, that is, to implement distributed inference. At present, the participants of vertical federated learning include two roles of server (Server) and client (Client). In addition, there is a more complex case of vertical federated learning participants, that is, a server and multiple clients. In this case, the distributed inference process needs the server and the multiple clients to participate together to complete.

[0089] Therefore, how to improve the feasibility of distributed inference in the vertical federated learning process is a problem to be solved.

[0090] To solve the above problems, the disclosure provides a model inference method, which can determine different relevant information of vertical federated learning inference for different clients, send first request information including the relevant information of vertical federated learning inference corresponding to each client to each client, so as to facilitate each client to perform inference, and improve the feasibility of distributed inference in the vertical federated learning inference process. That is, by determining the relevant information of vertical federated learning inference for each client in the plurality of clients, generating first request information corresponding to each client, and sending the first request information to each client, the vertical federated learning inference request is initiated to the plurality of clients, and the plurality of clients are interacted with, which helps the plurality of clients and the server to jointly participate in the vertical federated learning inference process, and improves the feasibility of distributed inference in the vertical federated learning process. Specifically as follows.

[0091] FIG. 1 shows a flowchart of some embodiments of the model inference method of the disclosure.

[0092] As shown in FIG. 1, the model inference method includes steps 110 to 120, which are executed by the server.

[0093] In step 110, the relevant information of vertical federated learning inference is determined for each client in the plurality of clients.

[0094] In some embodiments, the relevant information of VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, first sample indication information, and first feature indication information.

[0095] By determining different relevant information of vertical federated learning inference for different clients, it provides the possibility for the plurality of clients and the server to jointly participate in vertical federated learning, so that the clients can use the collected data samples to perform inference at various levels to obtain a model with better performance.

[0096] In some embodiments, the VFL association identifier is used to associate the process of VFL and / or the VFL model; the first sample indication information is used to indicate the data sample; and / or the first feature indication information is used to indicate the data feature.

[0097] The VFL association identifiers corresponding to different clients can be the same. The VFL association identifier can be used to know which VFL task and / or VFL model between the server and other clients participating in vertical federated learning, and in addition, the VFL association identifier can also be associated with the VFL association identifier in the vertical federated learning training process.

[0098] The process of associating the VFL with the VFL correlation identifier and / or the VFL model ensures the accuracy of the vertical federated learning inference process and reduces the possibility of confusion in the vertical federated learning inference process.

[0099] In some embodiments, the first sample indication information is used to indicate the data sample of the server and / or the client. The first sample indication information includes at least one of a user equipment identifier (UE ID), a generic public subscription identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier (Sample ID), and a group identifier (Group ID).

[0100] The first feature indication information is used to indicate the data feature used by the server and / or the data feature used by the client. The first feature indication information includes at least one of service experience information, quality of experience metrics (QoE metrics), voice quality information (e.g., MOS (Mean Opinion Score)), and a feature identifier (Feature ID).

[0101] The first sample indication information and the first feature indication information are used to inform the server or the client of the data sample and the data feature used by the other party, thereby reducing the possibility of resource waste caused by the use of the same data sample and data feature between the server and the plurality of clients.

[0102] In some embodiments, the information related to the VFL inference corresponding to each client further includes at least one of an interoperation indicator corresponding to each client, an analysis identifier, filter information, and a VFL identifier. The analysis identifier is used to identify a requested analysis or to identify that the requested analysis and / or the analysis result is generated in a VFL manner. The VFL identifier is used to indicate a model trained by VFL. The filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0103] The filter information can be used to filter the information as needed when the server and the client collect data samples subsequently.

[0104] For example, the relevant information of the vertical federated learning inference for each of the plurality of clients can be determined by the following method: dividing, according to a preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients, to determine the relevant information of the VFL inference for each client, wherein the preset rule includes at least one of an internal configuration of the server, an internal logic of the server, a network element type of each client, relevant information of each client obtained from other network elements in the process of finding the client, and information associated by each client in the training process. The internal configuration of the server and the internal logic of the server may, for example, be specified by an operator, in addition, the network element type of the client may, for example, be AF or NWDAF, the relevant information of each client may, for example, be the service range of the client, and the information associated by each client in the training process may, for example, be the data samples and / or data features possessed by the client, and the specific implementation is as follows.

[0105] According to the preset rule, the user equipment identifier set in the sample indication information corresponding to the plurality of clients is divided to determine the user equipment identifier set corresponding to each client in the first sample indication information for each client.

[0106] According to the preset rule, the sample identifier in the sample indication information corresponding to the plurality of clients is divided to determine the sample identifier corresponding to each client in the first sample indication information for each client.

[0107] According to the preset rule, the feature identifier in the feature indication information corresponding to the plurality of clients is divided to determine the feature identifier corresponding to each client in the first feature indication information for each client.

[0108] According to the preset rule, the quality of experience indicator information in the feature indication information corresponding to the plurality of clients is divided to determine the quality of experience indicator information corresponding to each client in the first feature indication information for each client.

[0109] According to the preset rule, the area of interest in the filter information corresponding to the plurality of clients is divided to determine the area of interest corresponding to each client in the filter information for each client.

[0110] According to the preset rule, the quality of experience indicator information in the filter information corresponding to the plurality of clients is divided to determine the quality of experience indicator information corresponding to each client in the filter information for each client.

[0111] In step 120, the first request information is sent to each client so that each client performs inference.

[0112] The service end initiates a vertical federated learning inference request to the client end, the service end interacts with multiple client ends to jointly complete a vertical federated learning inference process, and the problem of not supporting the vertical federated learning inference process of multiple client ends in the 5GC is solved.

[0113] In some embodiments, the VFL server (NWDAF or trusted AF) (i.e., the service end) determines and sends a VFL inference request / subscription (i.e., the first request information) to each VFL client (i.e., each client), wherein the VFL inference request / subscription includes a VFL association identifier (i.e., a VFL association identifier corresponding to each client).

[0114] In some embodiments, for each VFL client (NWDAF), the VFL server (NWDAF or trusted AF) sends a VFL inference subscription (Nnwdaf_VFLInference_Subscribe) or a VFL subscription request (Nnwdaf_VFLInference_Request) to each VFL client.

[0115] In some embodiments, for each VFL client (trusted AF), the VFL server (NWDAF) sends a VFL inference subscription (Nnwdaf_VFLInference_Subscribe) or a VFL subscription request (Nnwdaf_VFLInference_Request) to each VFL client.

[0116] In some embodiments, for each VFL client (non-trusted AF), the VFL server (NWDAF) sends a VFL inference subscription (Nnwdaf_VFLInference_Subscribe) or a VFL subscription request (Nnwdaf_VFLInference_Request) to the NEF serving the AF (i.e., each VFL client (non-trusted AF)).

[0117] In some embodiments, for each VFL client (non-trusted AF), the NEF converts an arbitrary internal identifier (i.e., an internal VFL association identifier) into an external identifier (i.e., an external VFL association identifier) and sends a VFL inference subscription (Nnwdaf_VFLInference_Subscribe) or a VFL subscription request (Nnwdaf_VFLInference_Request) to the VFL client (non-trusted AF).

[0118] In some embodiments, the VFL association identifier is used to indicate a VFL client that will use a pre-trained VFL local model associated with the VFL association identifier.

[0119] In some embodiments, the first sample indication information in the VFL inference request / subscription includes a target of the VFL inference (e.g., a UE ID).

[0120] In some embodiments, the VFL inference service can be provided by the NWDAF as a VFL client, and enables the VFL server as a consumer to request or subscribe / unsubscribe the VFL inference.

[0121] In some embodiments, the VFL inference service can be provided by the AF as a VFL client, and enables the VFL server as a consumer to request or subscribe / unsubscribe the VFL inference.

[0122] In some embodiments, the first request information includes related information of the VFL inference corresponding to each client.

[0123] In some embodiments, the server sends the first request information to each client according to the second request information sent by the receiving consumer NF (Network Function).

[0124] For example, the second request information is used to request an analysis result (a final inference result), the second request information includes at least one of an analysis identifier and a VFL indicator, the analysis identifier is used to identify a requested analysis and / or the analysis result is generated using a VFL manner, and the VFL indicator is used to indicate a model of VFL training.

[0125] In some embodiments, the analysis identifier (Analytics ID) can indicate a requested analysis service experience (Service Experience).

[0126] In some embodiments, the analysis result is in a form of an analysis ID output, includes at least one of various data of the analysis ID output, or includes at least one of quality of experience index information, latency information, and 5G service quality index.

[0127] In some embodiments, the second request information further includes filter information (Filter Information), wherein the filter information includes at least one of S-NSSAI (Single Network Slice Selection Assistance Information), AID (Application Identifier), AOI (Area of Interest), and DNAI (Data Network Access Identifier).

[0128] In some embodiments, the filter information can be used to filter information when collecting data at the subsequent server and client, for example, only collecting a single network slice selection assistance information.

[0129] For example, the consumer NF is an analytic logic function (AnLF), and the server is an application function (AF).

[0130] In the above embodiment, the relevant information of the longitudinal federated learning inference corresponding to each of the plurality of clients is determined, wherein the relevant information of the longitudinal federated learning inference of each client includes at least one of a longitudinal federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The longitudinal federated learning process of each client is identified by the longitudinal federated learning association identifier, and the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information including the relevant information of the longitudinal federated learning inference corresponding to each client is sent to each client to request each client to perform longitudinal federated learning inference, thereby improving the feasibility of distributed inference in the longitudinal federated learning process.

[0131] In some embodiments, the server also performs local VFL inference on the data in the server to obtain an intermediate result of the local VFL inference.

[0132] In some embodiments, the VFL server (i.e., the server) can collect its local data and generate an intermediate local inference result (i.e., an intermediate result of the local VFL inference).

[0133] In some embodiments, the data in the server can be, for example, data local to the server, data collected by the server, or both data local to the server and data collected by the server.

[0134] For example, the server collects data samples from the corresponding network element according to the first request information, so that the server performs local VFL inference on the data in the server to obtain an intermediate result of the local VFL inference.

[0135] In some embodiments, the data samples collected by the server from the corresponding network element can be, for example, quality of experience indicator information, location information, RAT (Radio Access Technology) access information, and PDU (Protocol Data Unit) session information.

[0136] In some embodiments, the server also receives the first response information sent by each client, and then generates the inference result according to the first response information sent by each client and / or the intermediate result of the local VFL inference.

[0137] In some embodiments, the VFL client (i.e., the client) sends the intermediate local result (i.e., the intermediate result of the local VFL inference) to the VFL server (i.e., the server).

[0138] In some embodiments, each VFL client (NWDAF) sends the VFL inference notification (Nnwdaf_VFLInference_Notify) or the VFL inference request response (Nnwdaf_VFLInference_Request response) to the VFL server (NWDAF or trusted AF).

[0139] In some embodiments, each VFL client (trusted AF) sends the VFL inference notification (Nnwdaf_VFLInference_Notify) or the VFL inference request response (Nnwdaf_VFLInference_Request response) to the VFL server (NWDAF).

[0140] In some embodiments, each VFL client (untrusted AF) sends the VFL inference notification (Nnwdaf_VFLInference_Notify) or the VFL inference request response (Nnwdaf_VFLInference_Request response) to the NEF.

[0141] In some embodiments, for each VFL client (untrusted AF), the NEF converts an arbitrary external identifier (i.e., external VFL association identifier) into an internal identifier (i.e., internal VFL association identifier), and sends the VFL inference notification (Nnwdaf_VFLInference_Notify) or the VFL inference request response (Nnwdaf_VFLInference_Request response) to the VFL server (NWDAF).

[0142] For example, the inference result can be generated by aggregating the first response information sent by each client and / or the intermediate result of the local VFL inference.

[0143] For example, the first response information corresponding to each client includes at least one of the intermediate result of the VFL inference corresponding to each client, the second sample indication information, the second feature indication information, and the VFL association identifier.

[0144] The intermediate result of the VFL inference corresponding to each client can be, for example, a gradient, a loss, or other forms of data.

[0145] The client sends the VFL association identifier to the server to inform the server that the first response information sent by the client belongs to which VFL inference process.

[0146] The second sample indication information is used to indicate the data samples of the server and / or the client, and the second sample indication information includes at least one of a user equipment identifier, a general public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0147] The second feature indication information is used to indicate the data features used by the server and / or the client, and the second feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

[0148] The second sample indication information and the second feature indication information are used to inform the server or the client of the data samples and the data features used by the other party, thereby reducing the possibility of resource waste caused by the use of the same data samples and data features between the server and the multiple clients.

[0149] In addition, the first response information further includes first validity time information used to indicate the validity time of the intermediate result of the VFL inference.

[0150] The first validity time information in the first response information is used to inform the server of the validity time of the intermediate result inferred by the client, which helps to ensure the efficiency of the vertical federated learning inference process and reduces the possibility of resource waste caused by the long time of the server without reacting to the first response information.

[0151] In some embodiments, when the server is a non-trusted AF, the interaction between the server and the client (other NF) within the 5GC needs to pass through the NEF (Network Exposure Function), which can avoid the exposure of information within the 5GC to the outside of the 5GC and improve the security of information interaction, thereby improving the security of the VFL inference. Therefore, the server can also receive the first response information sent by the NEF.

[0152] The interaction between the client and the non-trusted server is realized through the NEF, which realizes the joint inference of multiple clients in the vertical federated learning inference process on the basis of ensuring the security within the core network.

[0153] If the NEF has the aggregation capability, the first response information can be an aggregation of at least one of the intermediate result of the VFL inference corresponding to each client, the second sample indication information, the second feature indication information, and the VFL association identifier received by the NEF, and then the server generates the inference result according to the first response information and / or the intermediate result of the local VFL inference.

[0154] For example, the inference result is generated by aggregating the first response information aggregated by the NEF and / or the intermediate result of the local VFL inference.

[0155] If the NEF does not have the aggregation capability, the server can also receive the first response information sent by each client and forwarded by the network exposure function NEF, wherein the first response information corresponding to each client includes at least one of the intermediate result of the VFL inference corresponding to each client, the second sample indication information, the second feature indication information, and the VFL association identifier; and then the inference result is generated according to the first response information sent by each client and / or the intermediate result of the local VFL inference.

[0156] For example, the inference result is generated by aggregating the first response information sent by each client and / or the intermediate result of the local VFL inference.

[0157] In some embodiments, if the NEF does not have the aggregation capability, the NEF places the received first response information sent by each client in the same indication or the same signaling message to forward the first response information sent by each client to the server.

[0158] After the server receives the first response information, the server can send second response information to the consumer NF, wherein the second response information includes at least one of the analysis result and the second validity time information, the analysis result is generated according to the inference result, and the second validity time information is used to indicate the validity time of the inference result.

[0159] In some embodiments, the analysis result can be regarded as the inference result of the server. For example, the analysis result can be in the form of analysis ID output, and includes at least one of the data in the analysis ID output.

[0160] In some embodiments, the server is a network data analysis function NWDAF, each client is an application function AF, or the server and each client are different NWDAFs, or the server is an AF and each client is an NWDAF.

[0161] In the following, some embodiments are described to illustrate how to process the interactive information when the AF is a non-trusted AF.

[0162] In a case where the AF is a non-trusted AF, sending the first request information to each client comprises: sending the first request information to each client through a network exposure function (NEF), wherein the NEF stores a mapping relationship between an internal VFL association identifier and an external VFL association identifier, a mapping relationship between the first internal sample indication information and the first external sample indication information, and a mapping relationship between the first internal feature indication information and the first external feature indication information.

[0163] In addition, in a case where the server is a non-trusted AF, the VFL association identifier in the first request information is an external VFL association identifier, the first sample indication information comprises first external sample indication information, and / or the first feature indication information comprises first external feature indication information, the NEF maps the external VFL association identifier in the first request information to an internal VFL association identifier, maps the first external sample indication information to first internal sample indication information, and / or maps the first external feature indication information to first internal feature indication information.

[0164] In a case where multiple clients exist for a non-trusted AF, for each client that is a non-trusted AF, the VFL association identifier in the first request information is an internal VFL association identifier, the first sample indication information comprises first internal sample indication information, and / or the first feature indication information comprises first internal feature indication information, the NEF maps the first internal VFL association identifier in the first request information to a first external VFL association identifier, maps the first internal sample indication information to first external sample indication information, and / or maps the first internal feature indication information to first external feature indication information.

[0165] In some embodiments, in a case where multiple clients exist for a non-trusted AF, wherein the NEF stores a mapping relationship between an internal intermediate result and an external intermediate result, a mapping relationship between an internal VFL association identifier and an external VFL association identifier, a mapping relationship between second internal sample indication information and second external sample indication information, and a mapping relationship between second internal feature indication information and second external feature indication information.

[0166] For each client that is a non-trusted AF, the intermediate result of the VFL inference in the first response information is an external intermediate result, the VFL association identifier is an external VFL association identifier, the second sample indication information comprises second external sample indication information, and / or the second feature indication information comprises second external feature indication information, the NEF maps the external intermediate result in the first response information to an internal intermediate result, maps the external VFL association identifier to an internal VFL association identifier, maps the second external sample indication information to second internal sample indication information, and / or maps the second external feature indication information to second internal feature indication information.

[0167] In the case of a non-trusted AF on the service side, the intermediate result of the VFL inference in the first response information is an internal intermediate result, the VFL association identifier is an internal VFL association identifier, the second sample indication information includes second internal sample indication information, and / or the second feature indication information includes second internal feature indication information, and the NEF maps the internal intermediate result in the first response information to an external intermediate result, maps the internal VFL association identifier to an external VFL association identifier, maps the second internal sample indication information to second external sample indication information, and / or maps the second internal feature indication information to second external feature indication information.

[0168] Through the mapping of internal information and external information, the security of the core network is protected, the possibility of internal information leakage of the core network is reduced, and the security of the VFL inference process is improved.

[0169] FIG. 2 shows a flowchart of another embodiment of the inference method of the model of the present disclosure.

[0170] As shown in FIG. 2, the inference method of the model includes steps 210 to 230, and the inference method of the model is performed by the NEF.

[0171] In step 210, third request information sent by the service side is received.

[0172] For example, the third request information includes at least one of a VFL association identifier, sample indication information corresponding to a plurality of clients, and feature indication information.

[0173] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model; the sample indication information is used to indicate the data sample; and / or the feature indication information is used to indicate the data feature.

[0174] Through the NEF receiving the third request information of the service side, direct interaction between the service side and the client is avoided, the security of the core network is ensured, and the risk of internal information leakage of the core network is reduced.

[0175] In step 220, according to the third request information, relevant information of vertical federated learning inference is determined for each client in the plurality of clients.

[0176] By determining different relevant information of vertical federated learning inference for different clients, it provides the possibility for the plurality of clients and the service side to jointly participate in vertical federated learning, so that the clients can use the collected data samples to perform inference at various levels in the subsequent process to obtain a model with better performance.

[0177] In some embodiments, the information related to the VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, third sample indication information, and third feature indication information.

[0178] The third sample indication information is used to indicate data samples of the server and / or the client, and the third sample indication information includes at least one of a user equipment identifier, a general public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0179] The fourth sample indication information is used to indicate data samples of the server and / or the client, and the fourth sample indication information includes at least one of a user equipment identifier, a general public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0180] The third feature indication information is used to indicate data features used by the server and / or data features used by the client, and the third feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

[0181] The fourth feature indication information is used to indicate data features used by the server and / or data features used by the client, and the fourth feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

[0182] In some embodiments, the information related to the VFL inference corresponding to each client further includes at least one of an interoperation indicator corresponding to each client, an analysis identifier, filter information, and a VFL identifier, the analysis identifier is used to identify a requested analysis or identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL identifier is used to indicate a VFL trained model, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0183] For example, the information related to the VFL inference for each client can be determined by dividing at least one of the sample indication information, the feature indication information, and the filter information corresponding to multiple clients according to a preset rule, and the preset rule includes at least one of an internal configuration of the server, an internal logic of the server, a network element type of each client, information related to each client obtained from other network elements in a process of finding the client, and information associated with each client in a training process. The specific division method is as described in the inference method of the model on the server side, and will not be described here.

[0184] In some embodiments, if the NEF does not have the division capability, that is, the NEF cannot determine the related information of the vertical federated learning inference for each of the plurality of clients according to the third request information. At this time, the server can determine the related information of the vertical federated learning inference for each of the plurality of clients according to the third request information, that is, obtain the third sub-request information. At this time, the NEF is responsible for receiving the third sub-request information sent by the server, and then sending the third sub-request information to each client.

[0185] In addition, in the case of dividing the third request information through the server, the server will send multiple third sub-request information to the NEF, which will be forwarded to each client through the NEF respectively; in the case of dividing the third request information through the NEF, the server only needs to send one third request information to the NEF.

[0186] In step 230, the third sub-request information is sent to each client so that each client performs inference, wherein the third sub-request information includes the related information of the VFL inference corresponding to each client.

[0187] The interaction between the server and the client is realized through the NEF, which guarantees the security of the core network and reduces the risk of information leakage in the core network.

[0188] In some embodiments, the NEF also receives the third sub-response information sent by each client, wherein the third sub-response information includes at least one of the intermediate result of the VFL inference of each client, the fourth feature indication information, the fourth sample indication information and the VFL association identifier.

[0189] If the NEF has the aggregation capability, the NEF will also aggregate the third sub-response information to obtain the third response information, and then send the third response information to the server, so that the server generates the inference result according to the third response information and / or the intermediate result of the local VFL inference.

[0190] For example, the server can generate the inference result by aggregating the third response information and / or the intermediate result of the local VFL inference.

[0191] In some embodiments, the third response information further includes third valid time information for indicating the valid time of the intermediate result of the VFL inference.

[0192] If the NEF does not have the aggregation capability, the NEF will directly send the third sub-response information to the server, so that the server generates the inference result according to the third sub-response information and / or the intermediate result of the local VFL inference.

[0193] For example, the NEF can send multiple third sub-response information in the same indication or the same signaling message to the service end.

[0194] For example, the service end can aggregate the third sub-response information and / or the intermediate result of the local VFL inference to generate the inference result.

[0195] In the vertical federated learning process, the service end and the client can be the following combinations of network elements: the service end is a network data analysis function (NWDAF), and each client is an application function (AF); or the service end and each client are different NWDAFs; or the service end is an AF, and each client is an NWDAF.

[0196] In the case where the service end is a non-trusted AF or multiple clients exist non-trusted AFs, the NEF maps the information included in the third sub-request information to ensure the security of the core network and reduce the possibility of information leakage, wherein the NEF stores the mapping relationship between the internal VFL association identifier and the external VFL association identifier, the mapping relationship between the third internal sample indication information and the third external sample indication information, and the mapping relationship between the third internal feature indication information and the third external feature indication information.

[0197] How the NEF maps the third sub-request information can refer to the method of mapping the first request information in the inference method of the model on the service end side, which will not be repeated here.

[0198] In the case where the service end is a non-trusted AF or multiple clients exist non-trusted AFs, the NEF maps the information included in the third response information to ensure the security of the core network and reduce the possibility of information leakage, wherein the NEF stores the mapping relationship between the internal intermediate result and the external intermediate result, the mapping relationship between the internal VFL association identifier and the external VFL association identifier, the mapping relationship between the third internal sample indication information and the third external sample indication information, and the mapping relationship between the third internal feature indication information and the third external feature indication information.

[0199] How the NEF maps the third response information can refer to the method of mapping the first response information in the inference method of the model on the service end side, which will not be repeated here.

[0200] In the case that the service end is a non-trusted AF or multiple clients exist a non-trusted AF, the NEF maps the information included in the third sub-response information to guarantee the security of the core network and reduce the possibility of information leakage, wherein the NEF stores a mapping relationship between internal intermediate results and external intermediate results, a mapping relationship between internal VFL association identifiers and external VFL association identifiers, a mapping relationship between third internal sample indication information and third external sample indication information, and a mapping relationship between third internal feature indication information and third external feature indication information.

[0201] How the NEF maps the third sub-response information can refer to the method of mapping the first response information in the inference method of the model on the service end side, which will not be described herein again.

[0202] Through the mapping between the internal information and the external information, the security of the core network is protected, and the possibility of internal information leakage of the core network is reduced.

[0203] In the above embodiment, the related information of the vertical federated learning inference corresponding to each client in the multiple clients is determined, wherein the related information of the vertical federated learning inference of each client includes at least one of the vertical federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The vertical federated learning process of each client is identified by the vertical federated learning association identifier, and the data sample in the service end is indicated by the sample indication information, and the data feature in the service end and / or the data feature in each client is indicated by the feature indication information. Then the first request information including the related information of the vertical federated learning inference corresponding to each client is sent to each client to request each client to perform the vertical federated learning inference, thereby improving the feasibility of distributed inference in the vertical federated learning process.

[0204] FIG. 3 shows a flowchart of still some embodiments of the inference method of the model of the present disclosure.

[0205] As shown in FIG. 3, the inference method of the model includes steps 310 to 320, and the inference method of the model is performed by the client.

[0206] In step 310, the first request information sent by the service end is received. The first request information includes the related information of the VFL inference corresponding to the client, and the related information of the VFL inference corresponding to the client includes at least one of the VFL association identifier corresponding to the client, the first sample indication information, and the first feature indication information. The related information of the VFL inference corresponding to the client is divided by the service end from the related information of the VFL inference corresponding to multiple clients.

[0207] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model; the first sample indication information is used to indicate the data sample; and / or the first feature indication information is used to indicate the data feature.

[0208] In addition, the first sample indication information is used to indicate the data sample, wherein the first sample indication information comprises at least one of the user equipment identifier, the generic public user identifier GPSI, the subscription permanent identifier SUPI, the location information, the sample identifier, and the group identifier.

[0209] The second sample indication information is used to indicate the data sample, wherein the second sample indication information comprises at least one of the user equipment identifier, the generic public user identifier GPSI, the subscription permanent identifier SUPI, the location information, the sample identifier, and the group identifier.

[0210] The first feature indication information is used to indicate the data feature used by the server and / or the data feature used by the client, wherein the first feature indication information comprises at least one of the service experience information, the quality of experience indicator information, the voice quality information, and the feature identifier.

[0211] The second feature indication information is used to indicate the data feature used by the server and / or the data feature used by the client, wherein the second feature indication information comprises at least one of the service experience information, the quality of experience indicator information, the voice quality information, and the feature identifier.

[0212] For example, the related information of the VFL inference corresponding to the client further comprises at least one of the interoperation indicator corresponding to the client, the analysis identifier, the filter information, and the VFL identifier, the analysis identifier is used to identify the requested analysis or identify that the requested analysis and / or the analysis result is generated in the VFL manner, the VFL identifier is used to indicate the model of the VFL training, and the filter information comprises at least one of the single network slice selection assistance information, the application identifier, the region of interest, and the data network access identifier.

[0213] When the server is a non-trusted AF or the client exists a non-trusted AF, that is, the server and the client cannot directly interact with each other, the client can receive the first request information sent by the server through the NEF. When the NEF has the division capability, the NEF can divide the complete first request information sent by the server, and then forward the divided first request information to the client; when the NEF does not have the division capability, the NEF can directly receive the first request information divided by the server, and then forward the first request information to the client.

[0214] In step 320, the VFL inference is performed according to the first request information.

[0215] In some embodiments, the VFL inference according to the first request information also requires the client to collect data samples from the corresponding network element according to the first request information.

[0216] After the client performs the VFL inference according to the first request information, the first response information can be obtained, and at this time, the client can send the first response information to the server, wherein the first response information includes at least one of the intermediate result of the VFL inference corresponding to the client, the second sample indication information, the second feature indication information, and the VFL association identifier.

[0217] In some embodiments, the first response information further includes: first valid time information, used to indicate the valid time of the intermediate result of the VFL inference.

[0218] By sending the first response information to the server, it is helpful for the server to generate the inference result according to the first response information and / or the intermediate result of the local VFL inference.

[0219] In some embodiments, the server is a network data analysis function (NWDAF), and the client is an application function (AF), or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

[0220] In the case that the server is a non-trusted AF or the client exists a non-trusted AF, that is, the server and the client cannot directly interact with each other, the client can send the first response information to the NEF, so that the NEF forwards the first response information to the server. When the NEF has aggregation capability, the NEF can aggregate the first response information sent by the client, and then forwards the aggregated first response information to the server; when the NEF does not have aggregation capability, the NEF can place multiple first response information in the same indication or the same signaling message, and forwards to the server.

[0221] In addition, in the case that the server is a non-trusted AF or the client exists a non-trusted AF, that is, the server and the client cannot directly interact with each other, the first request information and the first response information can also be mapped through the NEF to ensure the security of the core network.

[0222] Firstly, the NEF stores the mapping relationship between the internal VFL association identifier and the external VFL association identifier, the mapping relationship between the first internal sample indication information and the first external sample indication information, the mapping relationship between the first internal feature indication information and the first external feature indication information, and the mapping relationship between the internal intermediate result and the external intermediate result.

[0223] How the NEF maps the first request information and the first response information can refer to the method of mapping the first request information and the first response information in the inference method of the model on the server side, which will not be described here.

[0224] In the above embodiment, the related information of the vertical federated learning inference corresponding to each of the plurality of clients is determined, wherein the related information of the vertical federated learning inference of each client includes at least one of a vertical federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The vertical federated learning process of each client is identified by the vertical federated learning association identifier, and the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information including the related information of the vertical federated learning inference corresponding to each client is sent to each client to request each client to perform vertical federated learning inference, thereby improving the feasibility of distributed inference in the vertical federated learning process.

[0225] FIG. 4 shows a flowchart of still other embodiments of the inference method of the model of the present disclosure.

[0226] As shown in FIG. 4, the inference method of the model includes steps 410 to 420, and the inference method of the model is performed by the client.

[0227] In step 410, third sub-request information sent by a network exposure function (NEF) is received, wherein the third sub-request information includes related information of a vertical federated learning (VFL) inference corresponding to the client, and the related information of the VFL inference corresponding to the client includes at least one of a VFL association identifier corresponding to the client, third sample indication information, and third feature indication information. The related information of the VFL inference corresponding to the client is divided for the client from a plurality of related information of VFL inferences corresponding to clients obtained by the NEF from a server.

[0228] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model; the third sample indication information is used to indicate the data sample; and / or the third feature indication information is used to indicate the data feature.

[0229] In addition, the third sample indication information is used to indicate the data sample, wherein the third sample indication information includes at least one of a user equipment identifier, a general public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier.

[0230] The fourth sample indication information is used to indicate the data sample, wherein the fourth sample indication information includes at least one of a user equipment identifier, a general public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier.

[0231] The third feature indication information is used for indicating the data features used by the service end and / or the data features used by the client, and the third feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier.

[0232] The fourth feature indication information is used for indicating the data features used by the service end and / or the data features used by the client, and the fourth feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier.

[0233] For example, the related information of the VFL inference corresponding to the client further comprises at least one of an interoperation indicator corresponding to the client, analysis identifier, filter information, and VFL indicator, the analysis identifier is used for identifying that the requested analysis is generated in a VFL mode or that the requested analysis and / or analysis result is generated in a VFL mode, the VFL indicator is used for indicating a model of VFL training, and the filter information comprises at least one of single network slice selection assistance information, application identifier, region of interest, and data network access identifier.

[0234] When the NEF has the division capability, the NEF can divide the complete third request information sent by the service end, and then forward the divided third sub-request information to the client; when the NEF does not have the division capability, the NEF can directly receive the third sub-request information divided by the service end, and forward to the client.

[0235] In step 420, the VFL inference is performed according to the third sub-request information.

[0236] In some embodiments, performing the VFL inference according to the third sub-request information further requires the client to obtain the data samples from the corresponding network element according to the third sub-request information.

[0237] After the client performs the VFL inference according to the third sub-request information, the third sub-response information can be obtained, and at this time, the client can send the third sub-response information to the service end, wherein the third sub-response information comprises at least one of the intermediate result of the VFL inference of the client, the fourth feature indication information, the fourth sample indication information, and the VFL association identifier, so that the NEF generates the response information according to the third sub-response information and sends the response information to the service end.

[0238] In some embodiments, the third sub-response information further comprises third valid time information used for indicating the valid time of the intermediate result of the VFL inference.

[0239] By sending the third sub-response information to the service end, it is helpful for the service end to generate the inference result according to the third sub-response information and / or the intermediate result of the local VFL inference.

[0240] In some embodiments, the server is a network data analytics function (NWDAF), and the client is an application function (AF), or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

[0241] When the NEF has aggregation capability, the NEF can aggregate the third sub-response information sent by the client and then forward the aggregated third response information to the server; when the NEF does not have aggregation capability, the NEF can place multiple third sub-response information in the same indication or the same signaling message and forward to the server.

[0242] In addition, in the case that the server is a non-trusted AF or the client is a non-trusted AF, that is, the server and the client cannot directly interact with each other, the third request information and the third sub-response information can also be mapped by the NEF to ensure the security of the core network, as follows.

[0243] First, the NEF stores the mapping relationship between the internal VFL association identifier and the external VFL association identifier, the mapping relationship between the third internal sample indication information and the third external sample indication information, the mapping relationship between the third internal feature indication information and the third external feature indication information, and the mapping relationship between the internal intermediate result and the external intermediate result.

[0244] How the NEF maps the third sub-request information can refer to the method of mapping the first request information in the inference method of the model on the server side, which will not be repeated here.

[0245] How the NEF maps the third sub-response information and / or the third response information can refer to the method of mapping the first response information in the inference method of the model on the server side, which will not be repeated here.

[0246] In the above embodiments, the related information of the vertical federated learning inference corresponding to each of the plurality of clients is determined, wherein the related information of the vertical federated learning inference of each client includes at least one of a vertical federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The vertical federated learning process of each client is identified by the vertical federated learning association identifier, the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information including the related information of the vertical federated learning inference corresponding to each client is sent to each client to request each client to perform vertical federated learning inference, thereby improving the feasibility of distributed inference in the vertical federated learning process.

[0247] FIG. 5 shows a schematic diagram of some embodiments of the server of the present disclosure.

[0248] As shown in FIG. 5, the server 50 includes a first determining unit 51 and a first sending unit 52.

[0249] The first determining unit 51 is configured to determine, for each of a plurality of clients, related information of vertical federated learning (VFL) inference, wherein the related information of VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, first sample indication information, and first feature indication information.

[0250] In some embodiments, the VFL association identifier is used to associate a process of VFL and / or a VFL model; the first sample indication information is used to indicate a data sample; and / or the first feature indication information is used to indicate a data feature.

[0251] In some embodiments, the first sample indication information is used to indicate a data sample; the second sample indication information is used to indicate a data sample; the first sample indication information includes at least one of a user equipment identifier, a generic public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier, and / or the second sample indication information includes at least one of a user equipment identifier, a generic public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier.

[0252] In some embodiments, the first feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client; the second feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client; the first feature indication information includes at least one of service experience information, quality of experience (QoE) indicator information, voice quality information, and a feature identifier, and / or the second feature indication information includes at least one of service experience information, quality of experience (QoE) indicator information, voice quality information, and a feature identifier.

[0253] In some embodiments, the related information of VFL inference corresponding to each client further includes at least one of an interoperation indicator corresponding to each client, an analysis identifier, filter information, and a VFL indicator, the analysis identifier is used to identify a requested analysis or identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a VFL trained model, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0254] In some embodiments, the first determining unit 51 is further configured to divide, according to a preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients, and determine, for each client, the related information of the VFL inference, wherein the preset rule includes at least one of an internal configuration of the server, an internal logic of the server, a network element type of each client, related information of each client obtained from other network elements in the process of finding the client, and information associated by each client in the training process.

[0255] In some embodiments, the first determining unit 51 is further configured to perform at least one of the following: dividing, according to a preset rule, a set of user equipment identifiers in the sample indication information corresponding to the plurality of clients, and determining, for each client, a set of user equipment identifiers corresponding to each client in the first sample indication information; dividing, according to a preset rule, a sample identifier in the sample indication information corresponding to the plurality of clients, and determining, for each client, a sample identifier corresponding to each client in the first sample indication information; dividing, according to a preset rule, a feature identifier in the feature indication information corresponding to the plurality of clients, and determining, for each client, a feature identifier corresponding to each client in the first feature indication information; dividing, according to a preset rule, quality of experience indicator information in the feature indication information corresponding to the plurality of clients, and determining, for each client, quality of experience indicator information corresponding to each client in the first feature indication information; dividing, according to a preset rule, a region of interest in the filter information corresponding to the plurality of clients, and determining, for each client, a region of interest corresponding to each client in the filter information; and dividing, according to a preset rule, quality of experience indicator information in the filter information corresponding to the plurality of clients, and determining, for each client, quality of experience indicator information corresponding to each client in the filter information.

[0256] The first sending unit 52 is configured to send the first request information to each client for inference, wherein the first request information includes the related information of the VFL inference corresponding to each client. In some embodiments, the server is a network data analysis function (NWDAF), and each client is an application function (AF), or the server and each client are different NWDAFs, or the server is an AF, and each client is an NWDAF.

[0257] In some embodiments, the first sending unit 52 is further configured to send, to each client, first request information through a network exposure function (NEF) in the case that the AF is a non-trusted AF, wherein in the case that the service end is a non-trusted AF, the VFL association identifier in the first request information is an external VFL association identifier, the first sample indication information includes first external sample indication information, and / or the first feature indication information includes first external feature indication information, the NEF maps the external VFL association identifier in the first request information to an internal VFL association identifier, maps the first external sample indication information to first internal sample indication information, and / or maps the first external feature indication information to first internal feature indication information, in the case that there are non-trusted AFs among the plurality of clients, for each client that is a non-trusted AF, the VFL association identifier in the first request information is an internal VFL association identifier, the first sample indication information includes first internal sample indication information, and / or the first feature indication information includes first internal feature indication information, the NEF maps the first internal VFL association identifier in the first request information to a first external VFL association identifier, maps the first internal sample indication information to first external sample indication information, and / or maps the first internal feature indication information to first external feature indication information.

[0258] In some embodiments, the service end 50 further includes a first collecting unit, wherein the first collecting unit is configured to collect data samples from corresponding network elements according to the first request information.

[0259] In some embodiments, the service end 50 further includes a third inference unit, wherein the third inference unit is configured to perform local VFL inference according to data in the service end to obtain an intermediate result of the local VFL inference.

[0260] In some embodiments, the service end 50 further includes a fourth receiving unit, wherein the fourth receiving unit is configured to receive first response information sent by each client, wherein the first response information corresponding to each client includes at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier; and generate an inference result according to the first response information sent by each client and / or the intermediate result of the local VFL inference.

[0261] In some embodiments, the first response information further includes first validity time information used to indicate a validity time of the intermediate result of the VFL inference.

[0262] In some embodiments, the fourth receiving unit is further configured to receive first response information sent by a network exposure function (NEF), wherein the first response information is obtained by the NEF aggregating at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier; and generate the inference result according to the first response information and / or an intermediate result of local VFL inference.

[0263] In some embodiments, the fourth receiving unit is further configured to receive first response information sent by each client through a network exposure function (NEF), wherein the first response information corresponding to each client includes at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier; and generate the inference result according to the first response information sent by each client and / or an intermediate result of local VFL inference.

[0264] For example, the server 50 further includes a first aggregation unit, wherein the first aggregation unit is configured to aggregate the first response information sent by each client and the intermediate result of local VFL inference to generate the inference result.

[0265] In some embodiments, the fourth receiving unit is further configured to receive second request information sent by a consuming network function (NF), wherein the second request information is used to request an analysis result, the second request information includes at least one of an analysis identifier and a VFL indicator, the analysis identifier is used to identify a requested analysis and / or that the analysis result is generated using a VFL manner, and the VFL indicator is used to indicate a model trained by VFL, and wherein sending the first request information to each client includes: sending the first request information to each client according to the second request information.

[0266] In some embodiments, the analysis result is in the form of an analysis ID output, includes at least one of various data of the analysis ID output, or includes at least one of quality of experience indicator information, latency information, and 5G quality of service indicators.

[0267] In some embodiments, the consuming NF is an analysis logic function (AnLF), and the server is an application function (AF).

[0268] In some embodiments, the second request information further includes filter information, wherein the filter information includes at least one of single network slice selection assistance information, application identifier, area of interest, and data network access identifier.

[0269] In some embodiments, the first sending unit 52 is further configured to send second response information to the consumer NF, wherein the second response information comprises at least one of an analysis result and second valid time information, the analysis result is generated according to the inference result, and the second valid time information is used to indicate a valid time of the inference result.

[0270] In some embodiments, the server 50 further comprises a first mapping unit, wherein the first mapping unit is configured to, in a case that there are non-trusted AFs in the plurality of clients, for each client that is a non-trusted AF, the intermediate result of the VFL inference in the first response information is an external intermediate result, the VFL association identifier is an external VFL association identifier, the second sample indication information comprises second external sample indication information, and / or the second feature indication information comprises second external feature indication information, map, by the NEF, the external intermediate result in the first response information to an internal intermediate result, map the external VFL association identifier to an internal VFL association identifier, map the second external sample indication information to second internal sample indication information, and / or map the second external feature indication information to second internal feature indication information, in a case that the server is a non-trusted AF, map, by the NEF, the internal intermediate result in the first response information to the external intermediate result, map the internal VFL association identifier to the external VFL association identifier, map the second internal sample indication information to the second external sample indication information, and / or map the second internal feature indication information to the second external feature indication information.

[0271] In the above embodiments, for each client in the plurality of clients, related information of a longitudinal federated learning inference corresponding to the client is determined, wherein the related information of the longitudinal federated learning inference of each client comprises at least one of a longitudinal federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The longitudinal federated learning process of each client is identified by the longitudinal federated learning association identifier, and the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information comprising the related information of the longitudinal federated learning inference corresponding to each client is sent to each client to request each client to perform the longitudinal federated learning inference, thereby improving the feasibility of distributed inference in the longitudinal federated learning process.

[0272] FIG. 6 shows a schematic diagram of some embodiments of the NEF of the present disclosure.

[0273] As shown in FIG. 6, the NEF 60 includes a first receiving unit 61, a second determining unit 62, and a second sending unit 63.

[0274] The first receiving unit 61 is configured to receive third request information sent by a server, wherein the third request information includes at least one of a VFL association identifier, sample indication information corresponding to a plurality of clients, and feature indication information.

[0275] In some embodiments, the VFL association identifier is used to associate a process of the VFL and / or a VFL model; the sample indication information is used to indicate a data sample; and / or the feature indication information is used to indicate a data feature.

[0276] In some embodiments, the related information of the VFL inference corresponding to each client further includes at least one of an interoperation indicator corresponding to each client, an analysis identifier, filter information, and a VFL indicator, the analysis identifier is used to identify a requested analysis or identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a model trained by VFL, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and data network access identification.

[0277] The second determining unit 62 is configured to determine, according to the third request information, related information of vertical federated learning VFL inference for each client in the plurality of clients, wherein the related information of the VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, third sample indication information, and third feature indication information.

[0278] In some embodiments, the third sample indication information is used to indicate a data sample; the fourth sample indication information is used to indicate a data sample; the third sample indication information includes at least one of a user equipment identifier, a generic public subscriber identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier, and / or the fourth sample indication information includes at least one of a user equipment identifier, a generic public subscriber identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0279] In some embodiments, the third feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client; the fourth feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client; the third feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier, and / or the fourth feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

[0280] In some embodiments, the second determining unit 62 is further configured to divide, according to a preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients, to determine, for each client, the related information of the VFL inference, wherein the preset rule includes at least one of an internal configuration of the server, an internal logic of the server, a network element type of each client, related information of each client obtained from other network elements in the process of finding the client, and information associated in the training process of each client.

[0281] In some embodiments, the second determining unit 62 is further configured to perform at least one of the following: dividing, according to a preset rule, a set of user equipment identifiers in the sample indication information corresponding to the plurality of clients, to determine, for each client, a set of user equipment identifiers corresponding to each client in the third sample indication information; dividing, according to a preset rule, a sample identifier in the sample indication information corresponding to the plurality of clients, to determine, for each client, a sample identifier corresponding to each client in the third sample indication information; dividing, according to a preset rule, a feature identifier in the feature indication information corresponding to the plurality of clients, to determine, for each client, a feature identifier corresponding to each client in the third feature indication information; dividing, according to a preset rule, quality of experience indicator information in the feature indication information corresponding to the plurality of clients, to determine, for each client, quality of experience indicator information corresponding to each client in the third feature indication information; dividing, according to a preset rule, a region of interest in the filter information corresponding to the plurality of clients, to determine, for each client, a region of interest corresponding to each client in the filter information; and dividing, according to a preset rule, quality of experience indicator information in the filter information corresponding to the plurality of clients, to determine, for each client, quality of experience indicator information corresponding to each client in the filter information.

[0282] The second sending unit 63 is configured to send, to each client, third sub-request information for each client to perform inference, wherein the third sub-request information includes the related information of the VFL inference corresponding to each client.

[0283] In some embodiments, the first receiving unit 61 is further configured to receive third sub-response information sent by each client, wherein the third sub-response information includes at least one of an intermediate result of the VFL inference of each client, fourth feature indication information, fourth sample indication information, and a VFL association identifier; in the case that the NEF has aggregation capability, the third sub-response information is aggregated to obtain third response information; and the third response information is sent to the server, so that the server generates an inference result according to the third response information and / or the intermediate result of the local VFL inference.

[0284] In some embodiments, the third response information further comprises third validity time information indicating a validity time of the intermediate result of the VFL inference.

[0285] In some embodiments, the second sending unit 63 is further configured to, in a case where the NEF does not have the aggregation capability, send the third sub-response information to the server, so that the server generates the inference result according to the third sub-response information and / or the intermediate result of the local VFL inference.

[0286] In some embodiments, the server is a network data analytics function (NWDAF), and each client is an application function (AF), or the server and each client are different NWDAFs, or the server is an AF, and each client is an NWDAF.

[0287] In some embodiments, the NEF further comprises a second mapping unit, wherein the second mapping unit is configured to, in a case where the server is a non-trusted AF, map, in the third sub-request information, an external VFL association identifier to an internal VFL association identifier, map third external sample indication information to third internal sample indication information, and / or map third external feature indication information to third internal feature indication information, in a case where there are non-trusted AFs among the plurality of clients, for each client that is a non-trusted AF, map, in the third sub-request information, an internal VFL association identifier to an external VFL association identifier, map third internal sample indication information to third external sample indication information, and / or map third internal feature indication information to third external feature indication information.

[0288] In some embodiments, the second mapping unit is further configured to, in the case that the plurality of clients exist untrusted AFs, for each client that is an untrusted AF, in the case that the intermediate result of the VFL inference in the third response information is an external intermediate result, the VFL association identifier is an external VFL association identifier, the fourth sample indication information comprises fourth external sample indication information, and / or the fourth feature indication information comprises fourth external feature indication information, map the external intermediate result in the third response information to an internal intermediate result, map the external VFL association identifier to an internal VFL association identifier, map the fourth external sample indication information to fourth internal sample indication information, and / or map the fourth external feature indication information to fourth internal feature indication information, in the case that the intermediate result of the VFL inference in the third response information is an internal intermediate result, the VFL association identifier is an internal VFL association identifier, the fourth sample indication information comprises fourth internal sample indication information, and / or the fourth feature indication information comprises fourth internal feature indication information, map the internal intermediate result in the third response information to an external intermediate result, map the internal VFL association identifier to an external VFL association identifier, map the fourth internal sample indication information to fourth external sample indication information, and / or map the fourth internal feature indication information to fourth external feature indication information.

[0289] In some embodiments, the second mapping unit is further configured to, in the case that the plurality of clients exist untrusted AFs, for each client that is an untrusted AF, in the case that the intermediate result of the VFL inference in the third sub-response information is an external intermediate result, the VFL association identifier is an external VFL association identifier, the fourth sample indication information comprises fourth external sample indication information, and / or the fourth feature indication information comprises fourth external feature indication information, map the external intermediate result in the third sub-response information to an internal intermediate result, map the external VFL association identifier to an internal VFL association identifier, map the fourth external sample indication information to fourth internal sample indication information, and / or map the fourth external feature indication information to fourth internal feature indication information, in the case that the intermediate result of the VFL inference in the third sub-response information is an internal intermediate result, the VFL association identifier is an internal VFL association identifier, the fourth sample indication information comprises fourth internal sample indication information, and / or the fourth feature indication information comprises fourth internal feature indication information, map the internal intermediate result in the third sub-response information to an external intermediate result, map the internal VFL association identifier to an external VFL association identifier, map the fourth internal sample indication information to fourth external sample indication information, and / or map the fourth internal feature indication information to fourth external feature indication information.

[0290] In the above embodiment, the related information of the vertical federated learning inference corresponding to each of the plurality of clients is determined, wherein the related information of the vertical federated learning inference of each client includes at least one of a vertical federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The vertical federated learning process of each client is identified by the vertical federated learning association identifier, and the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information including the related information of the vertical federated learning inference corresponding to each client is sent to each client to request each client to perform vertical federated learning inference, thereby improving the feasibility of distributed inference in the vertical federated learning process.

[0291] FIG. 7 shows a schematic diagram of some embodiments of the client of the present disclosure.

[0292] As shown in FIG. 7, the client 70 includes a second receiving unit 71 and a first inference unit 72.

[0293] The second receiving unit 71 is configured to receive the first request information sent by the server, wherein the first request information includes the related information of the VFL inference corresponding to the client, the related information of the VFL inference corresponding to the client includes at least one of a VFL association identifier corresponding to the client, first sample indication information, and first feature indication information, and the related information of the VFL inference corresponding to the client is divided by the server from the related information of the VFL inference corresponding to a plurality of clients.

[0294] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model; the first sample indication information is used to indicate the data sample; and / or the first feature indication information is used to indicate the data feature.

[0295] In some embodiments, the first sample indication information is used to indicate the data sample; the second sample indication information is used to indicate the data sample; the first sample indication information includes at least one of a user equipment identifier, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier, and / or the second sample indication information includes at least one of a user equipment identifier, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

[0296] In some embodiments, the first feature indication information is used to indicate the data features used by the server and / or the data features used by the client, and the second feature indication information is used to indicate the data features used by the server and / or the data features used by the client; the first feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier, and / or the second feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier.

[0297] In some embodiments, the related information of the VFL inference corresponding to the client further comprises at least one of an interoperation indicator corresponding to the client, an analysis identifier, filter information, and a VFL identifier, the analysis identifier is used to identify that the requested analysis or the requested analysis and / or the analysis result is generated in a VFL manner, the VFL identifier is used to indicate a model trained by VFL, and the filter information comprises at least one of single network slice selection assistance information, application identifier, region of interest, and data network access identifier.

[0298] The first inference unit 72 is configured to perform VFL inference according to the first request information.

[0299] In some embodiments, the client 70 further comprises a second collection unit, wherein the second collection unit is configured to collect data samples from the corresponding network element according to the first request information.

[0300] In some embodiments, the client 70 further comprises a third sending unit, wherein the third sending unit is configured to send first response information to the server, and the first response information comprises at least one of intermediate results of the VFL inference corresponding to the client, second sample indication information, second feature indication information, and VFL association identifier.

[0301] In some embodiments, the first response information further comprises first validity time information used to indicate the validity time of the intermediate results of the VFL inference.

[0302] In some embodiments, the server is a network data analysis function (NWDAF), and the client is an application function (AF), or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

[0303] In the above embodiments, the related information of the vertical federated learning inference corresponding to each of the plurality of clients is determined, wherein the related information of the vertical federated learning inference of each client includes at least one of a vertical federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The vertical federated learning process of each client is identified by the vertical federated learning association identifier, and the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information including the related information of the vertical federated learning inference corresponding to each client is sent to each client to request each client to perform vertical federated learning inference, thereby improving the feasibility of distributed inference in the vertical federated learning process.

[0304] FIG. 8 shows a schematic diagram of another embodiment of the client of the present disclosure.

[0305] As shown in FIG. 8, the client 80 includes a third receiving unit 81 and a second inference unit 82.

[0306] The third receiving unit 81 is configured to receive third sub-request information sent by a network exposure function (NEF), wherein the third sub-request information includes related information of a vertical federated learning (VFL) inference corresponding to the client, and the related information of the VFL inference corresponding to the client includes at least one of a VFL association identifier corresponding to the client, third sample indication information, and third feature indication information. The related information of the VFL inference corresponding to the client is divided for the client from a plurality of related information of VFL inferences corresponding to clients obtained by the NEF from a server.

[0307] In some embodiments, the VFL association identifier is used to associate the process of the VFL and / or the VFL model; the third sample indication information is used to indicate the data sample; and / or the third feature indication information is used to indicate the data feature.

[0308] In some embodiments, the third sample indication information is used to indicate the data sample; the fourth sample indication information is used to indicate the data sample; the third sample indication information includes at least one of a user equipment identifier, a general public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier, and / or the fourth sample indication information includes at least one of a user equipment identifier, a general public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier.

[0309] In some embodiments, the third feature indication information is used to indicate the data features used by the server and / or the data features used by the client; the fourth feature indication information is used to indicate the data features used by the server and / or the data features used by the client; the third feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier, and / or the fourth feature indication information comprises at least one of service experience information, quality of experience index information, voice quality information, and feature identifier.

[0310] In some embodiments, the related information of the VFL inference corresponding to the client further comprises at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator corresponding to the client, the analysis identifier is used to identify the requested analysis or identify that the requested analysis and / or the analysis result is generated in a VFL manner, the VFL indicator is used to indicate a model trained by VFL, and the filter information comprises at least one of single network slice selection assistance information, application identifier, region of interest, and data network access identifier.

[0311] The second inference unit 82 is configured to perform inference according to the third sub-request information.

[0312] In some embodiments, the client 80 further comprises a fourth sending unit, wherein the fourth sending unit is configured to send third sub-response information to the NEF, wherein the third sub-response information comprises at least one of the intermediate result of the VFL inference of the client, the fourth feature indication information, the fourth sample indication information, and the VFL association identifier, so that the NEF generates response information according to the third sub-response information and sends the response information to the server.

[0313] In some embodiments, the server is a network data analysis function (NWDAF), and the client is an application function (AF), or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

[0314] In the above embodiments, the related information of the vertical federated learning inference corresponding to each of the plurality of clients is determined, wherein the related information of the vertical federated learning inference of each client comprises at least one of a vertical federated learning association identifier corresponding to each client, sample indication information, and feature indication information. The vertical federated learning process of each client is identified by the vertical federated learning association identifier, the data samples in the server are indicated by the sample indication information, and the data features in the server and / or the data features in each client are indicated by the feature indication information. Then, the first request information comprising the related information of the vertical federated learning inference corresponding to each client is sent to each client to request each client to perform vertical federated learning inference, thereby improving the feasibility of distributed inference in the vertical federated learning process.

[0315] FIG. 9 shows a schematic diagram of some embodiments of the electronic device of the present disclosure.

[0316] As shown in FIG. 9, the electronic device 90 of this embodiment includes a memory 91 and a processor 92 coupled to the memory 91, the processor 92 being configured to perform the inference method of the model in any one of the preceding embodiments based on instructions stored in the memory 91.

[0317] The memory 91 may, for example, include system memory, fixed non-volatile storage media, etc. The system memory may, for example, store an operating system, application programs, a Boot Loader, and other programs, etc.

[0318] The electronic device 90 can further include an input / output interface 93, a network interface 94, a storage interface 95, etc. These interfaces 93, 94, 95, and the memory 91 and the processor 92 may, for example, be connected through a bus 96. Among them, the input / output interface 93 provides a connection interface for display, mouse, keyboard, touch screen, microphone, speaker, etc. input / output devices. The network interface 94 provides a connection interface for various networking devices. The storage interface 95 provides a connection interface for external storage devices such as SD cards and U disks.

[0319] FIG. 10 shows a schematic diagram of some embodiments of the inference method of the model of the present disclosure.

[0320] As shown in FIG. 10, the inference method of the model includes steps S11 to S17.

[0321] In step S11, the consumer NF sends second request information to the server to request a vertical federated learning inference process.

[0322] In some embodiments, the second request information includes at least one of an analytics identifier and a VFL indicator, the analytics identifier being used to identify a requested analytics and / or the analytics result being generated using a VFL manner, and the VFL indicator being used to indicate a VFL trained model.

[0323] In some embodiments, the second request information further includes filter information, wherein the filter information includes at least one of a single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

[0324] In step S12, the server sends first request information to each client.

[0325] The first request information in step S12 includes information related to the vertical federated learning inference determined by the server for each of the plurality of clients, wherein the information related to the VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, first sample indication information, and first feature indication information; and the first request information is sent to each client so that each client performs inference.

[0326] In step S13, the client collects data samples from the corresponding NF according to the received first request information to perform vertical federated learning inference.

[0327] In step S14, the server and the client perform VFL inference.

[0328] In step S14a, the client performs VFL inference according to the collected data samples.

[0329] In step S14b, the server performs VFL inference according to data in the server.

[0330] The data in the server may be, for example, data local to the server and / or data collected by the server.

[0331] In step S15, each client sends first response information to the server, wherein the first response information corresponding to each client includes at least one of an intermediate result of VFL inference corresponding to each client, second sample indication information, second feature indication information, and a VFL association identifier.

[0332] In step S16, the server aggregates the intermediate result of each client and / or the intermediate result of the server's inference to obtain second response information.

[0333] The second response information includes at least one of an analysis result generated according to the inference result and second valid time information indicating the valid time of the inference result.

[0334] In step S17, the second response information is sent to the consumption NF.

[0335] In the above embodiment, the server initiates a vertical federated learning inference request to the plurality of clients, interacts with the plurality of clients, and completes the vertical federated learning inference process, thereby solving the problem that multiple clients cannot jointly participate in the vertical federated learning inference process in 5GC.

[0336] FIG. 11 shows a schematic diagram of another embodiment of the inference method of the model of the present disclosure.

[0337] As shown in FIG. 11, the inference method of the model includes steps S21 to S28.

[0338] In step S21, the service end sends the third request information or the third sub-request information to the NEF.

[0339] When the NEF has the division capability and the division is performed by the NEF, the service end sends the third request information to the NEF, and only one piece of third request information needs to be sent.

[0340] If the NEF does not have the division capability, the service end divides the third request information to obtain multiple third sub-request information, and the service end sends the third sub-request information to the NEF, and multiple pieces of third sub-request information need to be sent to the NEF.

[0341] In step S22, the NEF maps the parameters, as follows.

[0342] If the division is performed by the NEF, the NEF needs to divide the third request information to obtain the third sub-request information, and map the third sub-request information; or map the third request information, and then divide the mapped third request information.

[0343] If the division is performed by the service end, the NEF receives the third sub-request information, and the NEF needs to map the third sub-request information.

[0344] In step S23, the NEF sends the mapped third sub-request information to each client end.

[0345] In step S24, the client end collects data samples from the corresponding BF according to the third sub-request information.

[0346] In step S25, the service end and the client end perform VFL inference.

[0347] In step S25a, the client end performs VFL inference according to the collected data samples.

[0348] In step S25b, the service end performs VFL inference according to the data in the service end. The data in the service end may be, for example, data local to the service end and / or data collected by the service end.

[0349] In step S26, the client end sends the intermediate result of its inference to the NEF through third sub-response information.

[0350] In step S27, the NEF maps the parameters, as follows.

[0351] If the aggregation is performed by the NEF, the NEF sends the third response information to the service end, wherein the NEF can first map the third sub-response information and then aggregate the mapped third sub-response information to obtain the third response information; or the NEF can first aggregate the third sub-response information and then map the aggregated third sub-response information to obtain the third response information.

[0352] If the aggregation is performed by the service end, the NEF only maps the third sub-response information to obtain the mapped third sub-response information, and then sends the mapped third sub-response information to the service end.

[0353] In step S28, the NEF sends the third response information or the mapped third sub-response information to the service end, and then the service end aggregates the third response information or the mapped third sub-response information with the intermediate result of reasoning of the service end to obtain the final reasoning result.

[0354] In the above embodiment, the interaction between the service end and the client is realized by the NEF, the service end initiates a vertical federated learning reasoning request to multiple clients through the NEF, interacts with the multiple clients, and completes the reasoning process of the vertical federated learning, thereby solving the problem that multiple clients cannot jointly participate in the vertical federated learning reasoning process in the 5GC. In addition, when the service end is a non-trusted AF, the security of the core network is also guaranteed, and the risk of information leakage in the core network is reduced.

[0355] FIG. 12 shows a schematic diagram of still other embodiments of the model reasoning method of the present disclosure.

[0356] As shown in FIG. 12, the model reasoning method includes steps S310 to S319.

[0357] As shown in FIG. 12, the service end is a NWDAF, and the client includes a NWDAF and an AF, wherein the AF is a non-trusted AF, that is, the client (AF) cannot directly interact with the service end, and the details are as follows.

[0358] In step S310, the service end sends first request information or first sub-request information to the NEF.

[0359] If the NEF performs the division, the service end sends the first request information to the NEF, and if the service end performs the division, the service end sends the first sub-request information to the NEF.

[0360] In step S311, the NEF performs parameter mapping, and the details are as follows.

[0361] If the division is performed by the NEF, the NEF needs to divide the first request information to obtain first sub-request information, and map the first sub-request information; or map the first request information, and then divide the mapped first request information.

[0362] If the division is performed by the service end, the NEF receives the first sub-request information, and the NEF needs to map the first sub-request information.

[0363] In step S312, the NEF sends the mapped first sub-request information to the client (AF).

[0364] In step S313, the service end directly sends the first sub-request information to the client (NWDAF), wherein the first sub-request information is information divided by the service end according to the first request information.

[0365] In step S314, the client (NWDAF) collects data samples from corresponding NFs according to the first sub-request information.

[0366] In step S315, the client (NWDAF), the client (AF) and the service end perform VFL inference.

[0367] In step S315a, the client (NWDAF) performs VFL inference according to the collected data samples.

[0368] In step S315b, the client (AF) performs VFL inference according to the data of the client (AF).

[0369] In step S315c, the service end performs VFL inference according to the data of the service end. The data in the service end may be, for example, data local to the service end and / or data collected by the service end.

[0370] In step S316, the client (AF) sends the intermediate result of its inference to the NEF through first sub-response information.

[0371] In step S317, the NEF maps parameters, as follows.

[0372] If the aggregation is performed by the NEF, the NEF sends the first response information to the service end, wherein the NEF can first map the first sub-response information, and then aggregate the mapped first sub-response information to obtain the first response information; or the NEF can first aggregate the first sub-response information, and then map the aggregated first sub-response information to obtain the first response information.

[0373] If the aggregation is performed by the server side, the NEF only maps the first sub-response information to obtain the mapped first sub-response information, and then sends the mapped first sub-response information to the server side.

[0374] In step S318, the NEF sends the first response information or the mapped first sub-response information to the server side.

[0375] In step S319, the client (NWDAF) sends the intermediate result of its reasoning to the server side through the first sub-response information, and then the server side aggregates the intermediate results of reasoning of the client (NWDAF), the client (AF) and the server side to obtain the final reasoning result.

[0376] In the above embodiment, the interaction between the server side and the client side is realized by the NEF, the server side initiates a vertical federated learning reasoning request to multiple clients through the NEF, interacts with multiple clients, and completes the vertical federated learning reasoning process, solving the problem that multiple clients cannot jointly participate in the vertical federated learning reasoning process in the 5GC. In addition, when there is a non-trusted AF in the client, the security of the core network is also guaranteed, and the risk of information leakage in the core network is reduced.

[0377] In some embodiments, a computer program product is protected, including a computer program or instructions, which are executed by a processor to implement the inference method of the model of the present disclosure. The computer program product includes a computer program carried on a computer readable medium, and the computer program includes program codes for executing the method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from the network through the communication device, or installed from the storage device, or installed from the ROM. When the computer program is executed by the CPU, the above-mentioned functions defined in the method of the embodiments of the present disclosure are executed.

[0378] In some embodiments, a computer program is protected, including instructions that, when executed by a processor, cause the processor to perform the inference method of the model of the present disclosure.

[0379] Those skilled in the art will appreciate that embodiments of the present disclosure can be provided as methods, systems, or computer program products. Therefore, the present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable non-transitory storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0380] So far, the inference method, the server, the NEF, the client, the computer readable storage medium and the computer program product of the model of the present disclosure have been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein according to the above description.

[0381] The methods and systems of the present disclosure can be implemented in a number of ways. For example, the methods and systems of the present disclosure can be implemented via software, hardware, firmware, or any combination of software, hardware, and firmware. The above described order of steps for the methods is merely for illustration, and the steps of the methods of the present disclosure are not limited to the above specifically described order, unless otherwise specifically stated. Furthermore, in some embodiments, the present disclosure can also be implemented as programs recorded in recording media, which include machine readable instructions for implementing the methods according to the present disclosure. Thus, the present disclosure also covers the recording media storing the programs for executing the methods according to the present disclosure.

[0382] Although some specific embodiments of the present disclosure have been described in detail by way of examples, those skilled in the art should understand that the above examples are merely for illustration, and are not intended to limit the scope of the present disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A reasoning method of a model, executed by a server, comprising: determining, for each of a plurality of clients, related information of a vertical federated learning (VFL) inference, wherein the related information of the VFL inference corresponding to each of the clients comprises at least one of a VFL association identifier, first sample indication information, and first feature indication information corresponding to each of the clients; sending, to each of the clients, first request information for each of the clients to perform inference, wherein the first request information comprises the related information of the VFL inference corresponding to each of the clients.

2. The reasoning method of claim 1, wherein: the VFL association identifier is used to associate a process of VFL and / or a VFL model; the first sample indication information is used to indicate a data sample; and / or the first feature indication information is used to indicate a data feature.

3. The inference method according to claim 1 or 2, wherein, the related information of the VFL inference corresponding to each of the clients further comprises at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator corresponding to each of the clients, the analysis identifier is used to identify a requested analysis or to identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a VFL trained model, and the filter information comprises at least one of a single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

4. The inference method of claim 3, wherein, the determining, for each of the plurality of clients, related information of a vertical federated learning (VFL) inference comprises: dividing, according to a preset rule, at least one of sample indication information, feature indication information, and filter information corresponding to the plurality of clients to determine the related information of the VFL inference for each of the clients, wherein the preset rule comprises at least one of an internal configuration of the server, an internal logic of the server, a network element type of each of the clients, related information of each of the clients obtained from other network elements in a client searching process, and information associated in a training process by each of the clients.

5. The inference method of claim 4, wherein, the dividing, according to a preset rule, at least one of sample indication information, feature indication information, and filter information corresponding to the plurality of clients to determine the related information of the VFL inference for each of the clients comprises at least one of: dividing, according to the preset rule, a set of user equipment identifiers in the sample indication information corresponding to the plurality of clients to determine, for each of the clients, a set of user equipment identifiers corresponding to each of the clients in the first sample indication information; dividing, according to the preset rule, a sample identifier in the sample indication information corresponding to the plurality of clients to determine, for each of the clients, a sample identifier corresponding to each of the clients in the first sample indication information; dividing, according to the preset rule, a feature identifier in the feature indication information corresponding to the plurality of clients to determine, for each of the clients, a feature identifier corresponding to each of the clients in the first feature indication information; and / or dividing, according to the preset rule, at least one of the filter information corresponding to the plurality of clients to determine, for each of the clients, at least one of a single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier corresponding to each of the clients in the filter information. According to the preset rule, the quality of experience indicator information in the feature indication information corresponding to the plurality of clients is divided, and the quality of experience indicator information corresponding to each client in the first feature indication information is determined for each client; According to the preset rule, the region of interest in the filter information corresponding to the plurality of clients is divided, and the region of interest corresponding to each client in the filter information is determined for each client; According to the preset rule, the quality of experience indicator information in the filter information corresponding to the plurality of clients is divided, and the quality of experience indicator information corresponding to each client in the filter information is determined for each client.

6. The inference method of any one of claims 1-5, further comprising: performing VFL inference according to data in the server to obtain an intermediate result of the local VFL inference.

7. The inference method of any one of claims 1-6, further comprising: receiving first response information sent by each client, wherein the first response information corresponding to each client includes at least one of the intermediate result of the VFL inference corresponding to each client, second sample indication information, second feature indication information, and the VFL association identifier; generating an inference result according to the first response information sent by each client and / or the intermediate result of the local VFL inference.

8. The inference method of any one of claims 1-7, further comprising: receiving first response information sent by a network exposure function (NEF), wherein the first response information is obtained by the NEF aggregating at least one of the intermediate result of the VFL inference corresponding to each client, second sample indication information, second feature indication information, and the VFL association identifier; generating an inference result according to the first response information and / or the intermediate result of the local VFL inference.

9. The inference method of any one of claims 1-8, further comprising: receiving second request information sent by a consumer network function (NF), wherein the second request information is used to request an analysis result, and the second request information includes at least one of an analysis identifier and a VFL indicator, the analysis identifier is used to identify a requested analysis and / or an analysis result generated using a VFL manner, and the VFL indicator is used to indicate a VFL trained model; wherein the sending of the first request information to each client comprises: sending the first request information to each client according to the second request information.

10. The inference method of claim 9, wherein, The second request information further includes filter information, wherein the filter information includes at least one of a single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

11. The inference method of claim 9 or 10, further comprising: sending second response information to the consumer NF, wherein the second response information comprises at least one of the analysis result generated according to the inference result and second validity time information indicating a validity time of the inference result.

12. The inference method of any one of claims 7 to 11, wherein, The first response information further comprises first validity time information indicating a validity time of the intermediate result of the VFL inference.

13. The inference method of any one of claims 7-12, wherein The first sample indication information is used to indicate a data sample. The second sample indication information is used to indicate a data sample. The first sample indication information comprises at least one of a user equipment identity, a generic public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier, and / or The second sample indication information comprises at least one of a user equipment identity, a generic public user identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier.

14. The inference method of any one of claims 7-13, wherein The first feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client. The second feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client. The first feature indication information comprises at least one of service experience information, quality of experience (QoE) information, voice quality information, and a feature identifier, and / or The second feature indication information comprises at least one of service experience information, quality of experience (QoE) information, voice quality information, and a feature identifier.

15. The inference method of any one of claims 1-14, further comprising: collecting data samples from corresponding network elements according to the first request information.

16. The inference method of any one of claims 9-15, wherein The analysis result is in the form of an analysis ID output, comprises at least one of data in the analysis ID output, or comprises at least one of quality of experience (QoE) information, latency information, and 5G quality of service (QoS) indicators.

17. The inferential method of any of claims 9-16 wherein, The consumer NF is an analytics logic function (AnLF), and the server is an application function (AF).

18. The inference method of any one of claims 1-17, wherein The server is a network data analytics function (NWDAF), and each client is an application function (AF), or the server and each client are different NWDAFs, or the server is an AF and each client is an NWDAF.

19. The inference method of claim 18, wherein, In a case where the AF is a non-trusted AF, the sending of the first request information to each client comprises: sending the first request information to each client via a network exposure function (NEF), wherein, in the case that the non-trusted AF exists in the plurality of clients, for each client that is a non-trusted AF, the VFL correlation identifier in the first request information is an internal VFL correlation identifier, the first sample indication information comprises first internal sample indication information, and / or the first feature indication information comprises first internal feature indication information, the NEF maps the first internal VFL correlation identifier in the first request information to a first external VFL correlation identifier, maps the first internal sample indication information to first external sample indication information, and / or maps the first internal feature indication information to first external feature indication information.

20. The inference method of claim 18 or 19, further comprising: receiving first response information sent by the each client and forwarded by a network exposure function (NEF), wherein the first response information corresponding to the each client comprises at least one of an intermediate result of the VFL inference corresponding to the each client, second sample indication information, second feature indication information, and the VFL correlation identifier; generating an inference result according to the first response information sent by the each client and / or an intermediate result of a local VFL inference.

21. The inference method of claim 20, further comprising: in the case that the non-trusted AF exists in the plurality of clients, for each client that is a non-trusted AF, the intermediate result of the VFL inference in the first response information is an external intermediate result, the VFL correlation identifier is an external VFL correlation identifier, the second sample indication information comprises second external sample indication information, and / or the second feature indication information comprises second external feature indication information, the NEF maps the external intermediate result in the first response information to an internal intermediate result, maps the external VFL correlation identifier to an internal VFL correlation identifier, maps the second external sample indication information to second internal sample indication information, and / or maps the second external feature indication information to second internal feature indication information, ​ In a case where the service end is the untrusted AF, the intermediate result of the VFL inference in the first response information is an internal intermediate result, the VFL association identifier is an internal VFL association identifier, the second sample indication information includes second internal sample indication information, and / or the second feature indication information includes second internal feature indication information, and the NEF maps the internal intermediate result in the first response information to the external intermediate result, maps the internal VFL association identifier to an external VFL association identifier, maps the second internal sample indication information to second external sample indication information, and / or maps the second internal feature indication information to second external feature indication information.

22. The inferential method of any of claims 7-21, wherein, The generating of the inference result according to the first response information sent by each client and / or the intermediate result of the local VFL inference includes: The first response information sent by each client and the intermediate result of the local VFL inference are aggregated to generate the inference result.

23. An inference method of a model, performed by a network exposure function (NEF), comprising: receiving third request information sent by a service end, wherein the third request information includes a VFL association identifier and at least one of sample indication information and feature indication information corresponding to a plurality of clients; determining, according to the third request information, for each client in the plurality of clients, related information of vertical federated learning (VFL) inference, wherein the related information of the VFL inference corresponding to each client includes at least one of a VFL association identifier corresponding to each client, third sample indication information, and third feature indication information; sending third sub-request information to each client for inference by each client, wherein the third sub-request information includes the related information of the VFL inference corresponding to each client.

24. The inference method of claim 23, wherein: the VFL association identifier is used to associate a process of VFL and / or a VFL model; the sample indication information is used to indicate a data sample; and / or the feature indication information is used to indicate a data feature.

25. The reasoning method according to claim 23 or 24, wherein, the related information of the VFL inference corresponding to each client further includes at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator corresponding to each client, the analysis identifier is used to identify a requested analysis or to identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a VFL trained model, and the filter information includes at least one of single network slice selection assistance information, an application identifier, a region of interest, and a data network access identifier.

26. The inferential method of claim 25, wherein, the determining, according to the third request information, for each client in the plurality of clients, related information of vertical federated learning (VFL) inference includes: According to a preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients is divided, and relevant information of the VFL inference for each client is determined, wherein the preset rule includes at least one of the following: an internal configuration of the server, an internal logic of the server, a network element type of each client, relevant information of each client obtained from other network elements in a process of searching for the client, and information associated in a training process of each client.

27. The inference method of claim 26, wherein, According to the preset rule, at least one of the sample indication information, the feature indication information, and the filter information corresponding to the plurality of clients is divided, and relevant information of the VFL inference for each client is determined, wherein the preset rule includes at least one of the following: an internal configuration of the server, an internal logic of the server, a network element type of each client, relevant information of each client obtained from other network elements in a process of searching for the client, and information associated in a training process of each client. According to the preset rule, a set of user equipment identifiers in the sample indication information corresponding to the plurality of clients is divided, and a set of user equipment identifiers corresponding to each client in the third sample indication information is determined for each client. According to the preset rule, a sample identifier in the sample indication information corresponding to the plurality of clients is divided, and a sample identifier corresponding to each client in the third sample indication information is determined for each client. According to the preset rule, a feature identifier in the feature indication information corresponding to the plurality of clients is divided, and a feature identifier corresponding to each client in the third feature indication information is determined for each client. According to the preset rule, quality of experience indicator information in the feature indication information corresponding to the plurality of clients is divided, and quality of experience indicator information corresponding to each client in the third feature indication information is determined for each client. According to the preset rule, a region of interest in the filter information corresponding to the plurality of clients is divided, and a region of interest corresponding to each client in the filter information is determined for each client. According to the preset rule, quality of experience indicator information in the filter information corresponding to the plurality of clients is divided, and quality of experience indicator information corresponding to each client in the filter information is determined for each client.

28. The inference method of any one of claims 23-27, further comprising: receiving third sub-response information sent by each client, wherein the third sub-response information includes at least one of an intermediate result of the VFL inference of each client, fourth feature indication information, fourth sample indication information, and the VFL association identifier; in a case where the NEF has aggregation capability, aggregating the third sub-response information to obtain third response information; sending the third response information to the server, so that the server generates an inference result according to the third response information and / or an intermediate result of a local VFL inference.

29. The inference method of claim 28, wherein, The third response information further includes third validity time information for indicating a validity time of the intermediate result of the VFL inference.

30. The reasoning method of any one of claims 23-29, wherein the third sample indication information is configured to indicate data samples; and the fourth sample indication information is configured to indicate data samples. The third sample indication information comprises at least one of a user equipment identity, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identity, a group identity, and / or The fourth sample indication information comprises at least one of a user equipment identity, a generic public user identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identity, a group identity.

31. The reasoning method of any one of claims 23-30, wherein the third feature indication information is configured to indicate data features used by the server and / or data features used by the clients; and the fourth feature indication information is configured to indicate data features used by the server and / or data features used by the clients. The third feature indication information comprises at least one of service experience information, quality of experience indicator information, voice quality information, a feature identity, and / or The fourth feature indication information comprises at least one of service experience information, quality of experience indicator information, voice quality information, a feature identity.

32. The reasoning method of any one of claims 23-31, wherein the server is a network data analytics function NWDAF and each client is an application function AF, or the server and each client are different NWDAFs, or the server is an AF and each client is a NWDAF.

33. The reasoning method of claim 32, further comprising: in a case that the server is the untrusted AF, the VFL association identifier in the third sub-request information is an external VFL association identifier, the third sample indication information comprises third external sample indication information, and / or the third feature indication information comprises third external feature indication information, mapping the external VFL association identifier in the third sub-request information to an internal VFL association identifier, mapping the third external sample indication information to third internal sample indication information, and / or mapping the third external feature indication information to third internal feature indication information, in a case that the multiple clients include the untrusted AF, for each client that is an untrusted AF, the VFL association identifier in the third sub-request information is an internal VFL association identifier, the third sample indication information comprises third internal sample indication information, and / or the third feature indication information comprises third internal feature indication information, mapping the internal VFL association identifier in the third sub-request information to an external VFL association identifier, mapping the third internal sample indication information to third external sample indication information, and / or mapping the third internal feature indication information to third external feature indication information.

34. The reasoning method of any one of claims 28-33, further comprising: ​ ​ ​ ​ ​ ​ In the case that the NEF does not have aggregation capability, the third sub-response information is sent to the service end, so that the service end generates inference results according to the third sub-response information and / or local VFL inference intermediate results.

35. The inference method of any one of claims 28-34, further comprising: In the case that the plurality of clients exist the untrusted AF, for each client that is an untrusted AF, the VFL inference intermediate result in the third sub-response information is an external intermediate result, the VFL association identifier is an external VFL association identifier, the fourth sample indication information includes fourth external sample indication information, and / or the fourth feature indication information includes fourth external feature indication information, the external intermediate result in the third response information is mapped to an internal intermediate result, the external VFL association identifier is mapped to an internal VFL association identifier, the fourth external sample indication information is mapped to fourth internal sample indication information, and / or the fourth external feature indication information is mapped to fourth internal feature indication information, In the case that the service end is the untrusted AF, the VFL inference intermediate result in the third sub-response information is an internal intermediate result, the VFL association identifier is an internal VFL association identifier, the fourth sample indication information includes fourth internal sample indication information, and / or the fourth feature indication information includes fourth internal feature indication information, the internal intermediate result in the third response information is mapped to the external intermediate result, the internal VFL association identifier is mapped to an external VFL association identifier, the fourth internal sample indication information is mapped to fourth external sample indication information, and / or the fourth internal feature indication information is mapped to fourth external feature indication information.

36. An inference method of a model, performed by a client, comprising: receiving first request information sent by a service end, wherein the first request information includes VFL inference related information corresponding to the client, the VFL inference related information corresponding to the client includes at least one of a VFL association identifier corresponding to the client, first sample indication information, and first feature indication information, and the VFL inference related information corresponding to the client is divided for the client by the service end from VFL inference related information corresponding to a plurality of clients; performing VFL inference according to the first request information.

37. The inference method of claim 36, wherein: the VFL association identifier is used to associate a process of VFL and / or a VFL model; the first sample indication information is used to indicate a data sample; and / or the first feature indication information is used to indicate a data feature.

38. The inference method of claim 36 or 37, further comprising: sending first response information to the service end, wherein the first response information includes at least one of a VFL inference intermediate result corresponding to the client, second sample indication information, second feature indication information, and the VFL association identifier.

39. The inferential method of any of claims 36-38, wherein, The related information of the VFL inference corresponding to the client further includes at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator corresponding to the client, the analysis identifier is used to identify a requested analysis or identify that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used to indicate a VFL trained model, and the filter information includes at least one of single network slice selection assistance information, application identification, a region of interest, and data network access identification.

40. The inferential method of claim 38 or 39, wherein, The first response information further includes first valid time information used to indicate a valid time of the VFL inference intermediate result.

41. The inference method of any one of claims 38-40, wherein, The first sample indication information is used to indicate a data sample. The second sample indication information is used to indicate a data sample. The first sample indication information includes at least one of a user equipment identifier, a generic public subscriber identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier, and / or The second sample indication information includes at least one of a user equipment identifier, a generic public subscriber identifier (GPSI), a subscription permanent identifier (SUPI), location information, a sample identifier, and a group identifier.

42. The inference method of any one of claims 38-41, wherein, The first feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client. The second feature indication information is used to indicate a data feature used by the server and / or a data feature used by the client. The first feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier, and / or The second feature indication information includes at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

43. The inference method of any one of claims 37-42, wherein, The server is a network data analytics function (NWDAF), and the client is an application function (AF), or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

44. An inference method of a model, performed by a client, comprising: receiving third sub-request information sent by a network exposure function (NEF), wherein the third sub-request information includes related information of a VFL inference corresponding to the client, the related information of the VFL inference corresponding to the client includes at least one of a VFL association identifier corresponding to the client, third sample indication information, and third feature indication information, and the related information of the VFL inference corresponding to the client is divided for the client from multiple pieces of related information of VFL inferences corresponding to clients acquired by the NEF from a server; performing inference according to the third sub-request information.

45. The inference method of claim 44, wherein, The VFL association identifier is used to associate a process of VFL and / or a VFL model. The third sample indication information is used for indicating data samples; and / or The third feature indication information is used for indicating data features.

46. The inference method of claim 44 or 45, further comprising: sending, to the NEF, third sub-response information, wherein the third sub-response information comprises at least one of an intermediate result of the VFL inference of the client, fourth feature indication information, fourth sample indication information, and the VFL association identifier, so that the NEF generates third response information according to the third sub-response information and sends the third response information to the server.

47. The inferential method of any of claims 44-46, wherein, The related information of the VFL inference corresponding to the client further comprises at least one of an interoperation indicator, an analysis identifier, filter information, and a VFL indicator corresponding to the client, the analysis identifier is used for identifying a requested analysis or identifying that a requested analysis and / or an analysis result is generated in a VFL manner, the VFL indicator is used for indicating a VFL trained model, and the filter information comprises at least one of single network slice selection assistance information, an application identifier, a region of interest, and data network access identifier.

48. The inference method of claim 46 or 47, wherein The third sample indication information is used for indicating data samples; The fourth sample indication information is used for indicating data samples; The third sample indication information comprises at least one of a user equipment identifier, a generic public subscriber identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier, and / or The fourth sample indication information comprises at least one of a user equipment identifier, a generic public subscriber identifier GPSI, a subscription permanent identifier SUPI, location information, a sample identifier, and a group identifier.

49. The inference method of any one of claims 46 to 48, wherein The third feature indication information is used for indicating data features used by the server and / or data features used by the client; The fourth feature indication information is used for indicating data features used by the server and / or data features used by the client; The third feature indication information comprises at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier, and / or The fourth feature indication information comprises at least one of service experience information, quality of experience indicator information, voice quality information, and a feature identifier.

50. The inference method of any one of claims 44 to 49, wherein The server is a network data analytics function NWDAF, and the client is an application function AF, or the server and the client are different NWDAFs, or the server is an AF, and the client is an NWDAF.

51. A server, comprising: a first determining unit configured to determine, for each client in a plurality of clients, related information of vertical federated learning VFL inference, wherein the related information of the VFL inference corresponding to each client comprises at least one of a VFL association identifier corresponding to the client, first sample indication information, and first feature indication information. The first sending unit is configured to send first request information to each client to perform inference, wherein the first request information comprises relevant information of the VFL inference corresponding to each client.

52. A NEF, comprising: The first receiving unit is configured to receive third request information sent by a server, wherein the third request information comprises at least one of a VFL association identifier, sample indication information corresponding to a plurality of clients, and feature indication information. The second determining unit is configured to determine, according to the third request information, relevant information of vertical federated learning (VFL) inference for each client in the plurality of clients, wherein the relevant information of the VFL inference corresponding to each client comprises at least one of a VFL association identifier corresponding to each client, third sample indication information, and third feature indication information. The second sending unit is configured to send third sub-request information to each client to perform inference, wherein the third sub-request information comprises relevant information of the VFL inference corresponding to each client.

53. A client, comprising: The second receiving unit is configured to receive first request information sent by a server, wherein the first request information comprises relevant information of the VFL inference corresponding to the client, the relevant information of the VFL inference corresponding to the client comprises at least one of a VFL association identifier corresponding to the client, first sample indication information, and first feature indication information, and the relevant information of the VFL inference corresponding to the client is divided for the client by the server from relevant information of VFL inference corresponding to a plurality of clients. The first inference unit is configured to perform VFL inference according to the first request information.

54. A client, comprising: The third receiving unit is configured to receive third sub-request information sent by a network exposure function (NEF), wherein the third sub-request information comprises relevant information of the VFL inference corresponding to the client, the relevant information of the VFL inference corresponding to the client comprises at least one of a VFL association identifier corresponding to the client, third sample indication information, and third feature indication information, and the relevant information of the VFL inference corresponding to the client is divided for the client by the NEF from relevant information of VFL inference corresponding to a plurality of clients obtained by the server. The second inference unit is configured to perform inference according to the third sub-request information.

55. An electronic device, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute an inference method according to any one of claims 1 to 50 based on instructions stored in the memory.

56. A computer-readable storage medium having stored thereon a computer program, the program being executed by a processor to implement an inference method according to any one of claims 1 to 50.

57. A computer program product comprising a computer program which, when executed by a processor, implements the inference method of any one of claims 1 to 50.

58. A computer program comprising: instructions which, when executed by the processor, cause the processor to perform the inference method of any one of claims 1 to 50.

Images