Support for joint learning (FL)

Abstract

The present disclosure relates to a server, a client, a network node and a method for supporting joint learning (FL). A method at a server associated with an FL process includes at least one of: sending a first message to one or more clients associated with the FL process, the first message indicating that the corresponding client is deselected by the server for the FL process and / or the FL process is suspended; and receiving a second message from one or more second network nodes associated with the FL process, the second message indicating that the corresponding client is exiting the FL process.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to PCT International Application No. PCT / CN2023 / 071204, filed on January 9, 2023, entitled “SUPPORT FOR FEDERATED LEARNING (FL)”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates to the field of telecommunications, and in particular, to a server, a client, a network node, and a method for supporting federated learning (FL). Background Art

[0004] With the advancement of electronics and telecommunications technology, mobile devices (e.g., mobile phones, smartphones, laptops, tablets, and in-car devices) have become an integral part of our daily lives. To support this large number of mobile devices, an efficient core network, such as the 3rd Generation Partnership Project (3GPP) Fifth Generation Core (5GC), will be required.

[0005] Machine learning (ML) is a key enabler for optimizing, securing, and managing core networks. This leads to increased collection and processing of data from network functions, which in turn can increase threats to sensitive end-user information. Therefore, mechanisms are needed to mitigate threats to end-user privacy and fully leverage the benefits of ML.

[0006] Federated learning (FL), also known as collaborative learning, is a new ML technique that trains algorithms across multiple decentralized edge devices or servers that store local data samples without exchanging those local data samples. This approach differs significantly from traditional centralized machine learning techniques (in which all local datasets are uploaded to a single server) and more classic decentralized approaches (which typically assume that local data samples are uniformly distributed).

[0007] Federated learning enables multiple actors to build a common robust machine learning model without sharing data, allowing key issues such as data privacy, data security, data access rights, and access to heterogeneous data to be addressed. Its applications span multiple industries including defense, telecommunications, the Internet of Things (IoT), and pharmaceuticals.

[0008] Therefore, it becomes increasingly meaningful to support FL in 3GPP 5GC. Summary of the Invention

[0009] According to the latest technical report from 3GPP, clause 8.8 of Technical Report (TR) 23.700-81 V2.0.0 (2022-11), the maintenance of the FL procedure between multiple network data analysis functions (NWDAFs) is added in the conclusion part of 5GC (mainly in Principle 5), as follows:

[0010] Principle 5: The NWDAF, including the Model Training Logic Function (MTLF) acting as a FL server, can determine the final list of NWDAFs including MTLFs acting as FL clients via the initial FL request to the FL client to determine the availability and compatibility of FL clients. During the FL process, based on local policies or the status of the FL client (e.g., load, availability, capabilities, latency, accuracy, etc.), the NWDAF, including the MTLF acting as a FL server, can trigger the reselection, addition, or removal of FL clients and can publish new FL client discovery via the Network Repository Function (NRF). FL clients can dynamically join or exit FL operations during the execution phase.

[0011] However, for the maintenance and implementation of the FL process, it remains unclear how the server NWDAF obtains the analytical data (e.g., network function (NF) load, etc.) of the client NWDAF and how to terminate the federated learning operation at the client NWDAF.

[0012] In order to solve or at least alleviate the above problems, some embodiments of the present disclosure provide a server, a client, a network node, and a method for supporting federated learning (FL) in a core network (eg, a 5GC network).

[0013] According to a first aspect of the present disclosure, a method is provided at a server associated with a FL process. The method includes at least one of the following: sending a first message to one or more clients associated with the FL process, the first message indicating that the corresponding client has been deselected by the server for the FL process and / or that the FL process is suspended; and receiving a second message from the one or more clients associated with the FL process, the second message indicating that the corresponding client is exiting the FL process. In some embodiments, the method further includes any of the steps of any method of the second aspect.

[0014] According to a second aspect of the present disclosure, a method is provided at a server associated with a FL process. The method includes: receiving one or more fifth messages indicating analysis data associated with one or more clients in the FL process and / or one or more candidate clients to be selected for the FL process; and selecting at least one client from the one or more candidate clients and / or the one or more clients for the FL process based at least on the analysis data. In some embodiments, the method further includes any of the steps of any of the methods of the first aspect.

[0015] According to a third aspect of the present disclosure, a server is provided, comprising: a processor; and a memory storing instructions, which, when executed by the processor, cause the processor to perform any method of the first aspect and / or the second aspect.

[0016] According to a fourth aspect of the present disclosure, a server associated with a FL process is provided. The server includes at least one of the following: a sending module configured to send a first message to one or more clients associated with the FL process, the first message indicating that the corresponding client has been deselected from the FL process by the server and / or that the FL process is suspended; and a receiving module configured to receive a second message from the one or more clients associated with the FL process, the second message indicating that the corresponding client is exiting the FL process. In some embodiments, the server includes one or more additional modules, each of which can perform any step of any method of the first aspect.

[0017] According to a fifth aspect of the present disclosure, a server associated with a FL process is provided. The server includes: a receiving module configured to receive one or more fifth messages indicating analysis data associated with one or more clients in the FL process and / or one or more candidate clients to be selected for the FL process; and a selection module configured to select at least one client from the one or more candidate clients and / or the one or more clients for the FL process based at least on the analysis data. In some embodiments, the server includes one or more additional modules, each of which can perform any step of any method of the second aspect.

[0018] According to a sixth aspect of the present disclosure, a method is provided at a client associated with a FL process. The method includes at least one of the following: receiving a first message from a server associated with the FL process, the first message indicating that the client has been deselected from the FL process by the server and / or that the FL process is suspended; and sending a second message to the server associated with the FL process, the second message indicating that the client is exiting the FL process. In some embodiments, the method further includes any of the steps of any of the methods of the seventh aspect.

[0019] According to a seventh aspect of the present disclosure, a method is provided at a client associated with a FL process or at a candidate client to be selected for the FL process. The method includes at least one of the following: sending a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with the client or candidate client; and sending a seventh message to one or more network nodes, the seventh message indicating data associated with the client or candidate client, the data used as input data when determining the analysis data associated with the client or candidate client. In some embodiments, the method also includes any of the steps of any of the methods of the sixth aspect.

[0020] According to an eighth aspect of the present disclosure, a client is provided, comprising: a processor; and a memory storing instructions, wherein when the processor executes the instructions, the processor executes any of the methods of the sixth and / or seventh aspects.

[0021] According to a ninth aspect of the present disclosure, a client associated with a FL process is provided. The client includes at least one of the following: a receiving module configured to receive a first message from a server associated with the FL process, the first message indicating that the client has been deselected from the FL process by the server and / or that the FL process is suspended; and a sending module configured to send a second message to the server associated with the FL process, the second message indicating that the client is exiting the FL process. In some embodiments, the client includes one or more additional modules, each of which can perform any step of any method of the sixth aspect.

[0022] According to a tenth aspect of the present disclosure, a client associated with a FL process or a candidate client to be selected for the FL process is provided. The client or candidate client includes at least one of the following: a first sending module configured to send a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with the client or candidate client; and a second sending module configured to send a seventh message to one or more network nodes, the seventh message indicating data associated with the client or candidate client, the data used as input data when determining the analysis data associated with the client or candidate client. In some embodiments, the client or candidate client includes one or more additional modules, each of which can perform any step of any method of the seventh aspect.

[0023] According to an eleventh aspect of the present disclosure, a method at a network node is provided, the method comprising: sending a fifth message to a server associated with a FL process, the fifth message indicating analysis data associated with one or more clients associated with the FL process and / or one or more candidate clients to be selected by the server for the FL process.

[0024] According to a twelfth aspect of the present disclosure, a network node is provided, comprising: a processor; and a memory storing instructions, wherein when the processor executes the instructions, the processor executes any method of the eleventh aspect.

[0025] According to a thirteenth aspect of the present disclosure, a network node is provided. The network node includes: a sending module configured to send a fifth message to a server associated with a FL process, the fifth message indicating analysis data associated with one or more clients associated with the FL process and / or one or more candidate clients to be selected by the server for the FL process. In some embodiments, the network node includes one or more additional modules, each of which can perform any step of any method of the eleventh aspect.

[0026] According to a fourteenth aspect of the present disclosure, there is provided a computer program comprising instructions, which, when executed by at least one processor, cause the at least one processor to perform any method of any one of the first, second, sixth, seventh and / or eleventh aspects.

[0027] According to a fifteenth aspect of the present disclosure, there is provided a carrier comprising the computer program of the fourteenth aspect. In some embodiments, the carrier is one of an electric signal, an optical signal, a radio signal or a computer-readable storage medium.

[0028] According to a sixteenth aspect of the present disclosure, a telecommunications system for supporting FL is provided. The telecommunications system comprises: a server according to the third, fourth, and / or fifth aspects; and one or more clients according to the eighth, ninth, and / or tenth aspects. In some embodiments, the telecommunications system further comprises one or more network nodes according to the twelfth and / or thirteenth aspects.

[0029] According to some embodiments of the present disclosure, FL can be supported in a core network, such as a 5GC network. This allows for the maintenance and implementation of a federated learning process. In one aspect, the server NWDAF can obtain analysis data from the client NWDAF via an auxiliary NWDAF and / or directly from the client NWDAF. Furthermore, in another aspect, the federated learning operation can terminate at the client NWDAF, while considering two different approaches for exchanging ML model information during the FL execution phase: reusing existing services (or extensions thereof) and using new services. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The foregoing features and other features of the present disclosure will become more fully apparent from the following description and appended claims taken in conjunction with the accompanying drawings. It should be understood that these drawings depict only several embodiments of the present disclosure and are therefore not to be considered as limiting the scope of the present disclosure, which will be described with additional specificity and detail through the use of the accompanying drawings.

[0031] Figure 1 is a diagram illustrating an exemplary telecommunication network in which support for FL is applicable according to some embodiments of the present disclosure.

[0032] Figure 2 is a diagram illustrating an exemplary process for client-side NWDAF selection in the FL preparation phase (where support for FL is applicable) according to some embodiments of the present disclosure.

[0033] Figure 3 is a diagram illustrating an exemplary process for NWDAF monitoring and reselection in the FL execution phase (where support for FL is applicable) according to some embodiments of the present disclosure.

[0034] Figure 4 is a diagram illustrating an exemplary process of dynamically discovering a new NWDAF in the FL execution phase when information about the server NWDAF is known at the new client NWDAF (where support for FL is applicable), according to some embodiments of the present disclosure.

[0035] Figure 5 is a diagram illustrating an exemplary process of dynamically discovering a new NWDAF in the FL execution phase when information about the server NWDAF is unknown at the new client NWDAF (where support for FL is applicable) according to some embodiments of the present disclosure.

[0036] Figure 6 is a diagram illustrating an exemplary system for analyzing data collection according to some embodiments of the present disclosure.

[0037] Figure 7is a diagram illustrating an exemplary system for terminating the FL process at a client NWDAF according to some embodiments of the present disclosure.

[0038] Figure 8 is a diagram illustrating an exemplary scenario for analyzing data collection according to some embodiments of the present disclosure.

[0039] Figure 9 is a diagram illustrating an exemplary scenario for terminating the FL process at the client NWDAF according to some embodiments of the present disclosure.

[0040] Figure 10 is a diagram illustrating an exemplary process for analyzing data collection according to some embodiments of the present disclosure.

[0041] Figure 11A and Figure 11B is a diagram illustrating an exemplary process for terminating the FL process at the client NWDAF according to some embodiments of the present disclosure.

[0042] Figure 12 is a flow chart illustrating an exemplary method at a server according to some embodiments of the present disclosure.

[0043] Figure 13 is a flowchart illustrating an exemplary method at a server according to another embodiment of the present disclosure.

[0044] Figure 14 is a flowchart illustrating an exemplary method at a client according to some embodiments of the present disclosure.

[0045] Figure 15 is a flowchart of another exemplary method at a client according to another embodiment of the present disclosure.

[0046] Figure 16 is a flowchart of an exemplary method at a network node according to an embodiment of the present disclosure.

[0047] Figure 17 Embodiments of arrangements that may be used in a server, a client and / or a network node according to embodiments of the present disclosure are schematically shown.

[0048] Figure 18 is a block diagram of an exemplary server according to an embodiment of the present disclosure.

[0049] Figure 19 is a block diagram of another exemplary server according to another embodiment of the present disclosure.

[0050] Figure 20 is a block diagram of an exemplary client according to an embodiment of the present disclosure.

[0051] Figure 21 is a block diagram of another exemplary client according to another embodiment of the present disclosure.

[0052] Figure 22 is a block diagram of an exemplary network node according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0053] Hereinafter, the present disclosure will be described with reference to the embodiments shown in the accompanying drawings. However, it should be understood that these descriptions are provided for illustrative purposes only and are not intended to limit the present disclosure. In addition, descriptions of known structures and technologies are omitted below so as not to unnecessarily obscure the concept of the present disclosure.

[0054] Those skilled in the art will understand that the term "exemplary" is used herein to mean "illustrative" or "serving as an example" and is not intended to imply that a particular embodiment is superior to another or that a particular feature is essential. Similarly, unless the context clearly indicates otherwise, the terms "first" and "second" and similar terms are used only to distinguish one particular instance of an item or feature from another particular instance, and do not indicate a particular order or arrangement. In addition, as used herein, the term "step" is intended to be synonymous with "operation" or "action." Unless the context or details of the operations being described clearly indicate otherwise, any description of a sequence of steps herein does not imply that the operations must be performed in a particular order, or even that the operations be performed in any order.

[0055] Unless expressly limited herein or understood from the context, conditional language (e.g., "can," "might," "could," "for example," etc.) as used herein is generally intended to convey that some embodiments include certain features, elements, and / or steps while other embodiments do not. Thus, such conditional language is generally not intended to imply that a feature, element, and / or step is always required for one or more embodiments, or that one or more embodiments must include logic to determine, with or without author input or permission, whether to include or implement such features, elements, and / or steps in any particular embodiment. Furthermore, the term "or" is used in an inclusive sense (and not an exclusive sense), such that when used, for example, to connect a list of elements, the term "or" means one, some, or all of the elements in the list. Furthermore, in addition to having its ordinary meaning, the term "each" as used herein may also refer to any subset of the set of elements to which the term "each" applies.

[0056] The term "based on" should be interpreted as "based, at least in part, on." The terms "one embodiment" and "an embodiment" should be interpreted as "at least one embodiment." The term "another embodiment" should be interpreted as "at least one other embodiment." The following may include other definitions, both explicit and implicit. Furthermore, unless otherwise expressly limited herein, statements such as "at least one of X, Y, and Z" should be understood in context as generally used to convey that an item, term, etc. can be X, Y, or Z, or a combination thereof.

[0057] The terms used herein are only used for the purpose of describing specific embodiments and are not intended to limit the example embodiments. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when used in this article, the words "comprising", "having", "including" indicate the presence of stated features, elements and / or components, etc., but do not exclude the presence or addition of one or more other features, elements, components and / or combinations thereof. It will also be understood that: unless clearly stated to the contrary, when used in this article, the terms "connected", "connected to", "connected to", etc. only mean that there is an electrical connection or communication connection between two elements and that they can be connected directly or indirectly.

[0058] Of course, without departing from the scope and essential features of the present disclosure, the present disclosure may be implemented in other specific ways than those set forth herein. One or more of the specific processes discussed below may be performed in any electronic device that includes one or more appropriately configured processing circuits, which, in some embodiments, may be embodied in one or more application-specific integrated circuits (ASICs). In some embodiments, these processing circuits may include one or more microprocessors, microcontrollers, and / or digital signal processors that are programmed with appropriate software and / or firmware to implement one or more of the above-described operations and variations thereof. In some embodiments, these processing circuits may include custom hardware that performs one or more of the functions described above. The embodiments presented are therefore to be considered in all respects as illustrative and not restrictive.

[0059] Although several embodiments of the present disclosure will be shown in the drawings and described in the following detailed description, it will be understood that the disclosure is not limited to the disclosed embodiments, but is capable of various rearrangements, modifications and substitutions without departing from the present disclosure as will be set forth and defined in the appended claims.

[0060] Furthermore, please note that while some embodiments of the present disclosure are described below in the context of the 5G system (5GS), the present disclosure is not limited thereto. In fact, the inventive concepts of the present disclosure are applicable to any appropriate communication architecture, such as Global System for Mobile Communications (GSM) / General Packet Radio Service (GPRS), Enhanced Data Rates for GSM Evolution (EDGE), Code Division Multiple Access (CDMA), Wideband CDMA (WCDMA), Time Division Synchronous CDMA (TD-SCDMA), CDMA2000, Worldwide Interoperability for Microwave Access (WiMAX), Wireless Fidelity (Wi-Fi), Universal Terrestrial Radio Access Network (UTRAN), Evolved UTRAN (E-UTRAN), Long Term Evolution (LTE), Evolved Packet System (EPS), and the like, as long as support for FL is involved. Therefore, those skilled in the art will readily appreciate that the terms used herein may also refer to their equivalents in any other infrastructure. For example, the term "user equipment" or "UE" as used herein may refer to a mobile device, mobile terminal, mobile station, user device, user terminal, wireless device, wireless terminal, IoT device, vehicle, or any other equivalent. As another example, the term "network node" as used herein may refer to or include a base station, a base transceiver station, an access point, a hotspot, a Node B (NB), an evolved Node B (eNB), a gNB, a network element, a network function, or any other equivalent.

[0061] In addition, the following 3GPP documents are incorporated herein by reference in their entirety:

[0062] 3GPP Technical Specification (TS) 23.288 V18.0.0 (2022-12), TechnicalSpecification, 3rd Generation Partnership Project;Technical SpecificationGroup Services and System Aspects;Architecture enhancements for 5G System(5GS) to support network data analytics services (Release 18));

[0063] 3GPP TR 23.700-81 V2.0.0 (2022-11), Technical Report, 3rd GenerationPartnership Project; Technical Specification Group Services and SystemAspects; Study of Enablers for Network Automation for 5G, 5G System (5GS); Phase 3 (Release 18).

[0064] Figure 1 1 is a block diagram illustrating an exemplary telecommunication network 10 in which support for FL is applicable according to an embodiment of the present disclosure. Although the telecommunication network 10 is a network defined in the context of 5GS, the present disclosure is not limited thereto.

[0065] like Figure 1 As shown, the network 10 may include one or more UEs 100 and a (radio) access network ((R)AN) 105, which includes one or more RAN nodes, such as base stations, Node Bs, evolved NodeBs (eNBs), gNBs, or access network (AN) nodes, which provide the UEs 100 with access to other parts of the network 10. In addition, the network 10 may include a core network portion thereof including, but not limited to, one or more user plane functions (UPFs) 115, NWDAFs 120, authentication server functions (AUSFs) 125, access and mobility management functions (AMFs) 130, session management functions (SMFs) 135, service communication brokers (SCPs) 140, network slice admission control functions (NSACFs) 145, network slice selection functions (NSSFs) 150, network exposure functions (NEFs) 155, network repository functions (NRFs) 160, policy control functions (PCFs) 165, unified data management (UDM) 170, application functions (AFs) 175, and edge application server discovery functions (EASDFs) 180. Figure 1 As shown, these entities can communicate with each other via service-based interfaces (e.g., Namf, Nsmf, etc.) and / or reference points (e.g., N1, N2, N3, N4, N6, N9, etc.).

[0066] However, the present disclosure is not limited thereto. In some other embodiments, the network 10 may include additional network functions, fewer network functions, or Figure 1Some variations of the existing network functions shown. For example, in a network with a 4G EPS architecture, the entity that performs these functions (e.g., the Mobility Management Entity (MME)) can be connected to Figure 1 For another example, in a network with a hybrid 4G / 5G architecture, some of these entities may be different from the entities shown in FIG. Figure 1 Some entities shown are identical, while others may be different. Figure 1 The functions shown are not essential to the embodiments of the present disclosure. In other words, some of them may be omitted from some embodiments of the present disclosure.

[0067] like Figure 1 As shown, the UPF 115 is communicatively connected to a data network (DN) 185, which may be the Internet or is in turn communicatively connected to the Internet, so that the UE 100 can ultimately communicate its user plane data with other devices outside the network 10, for example, via the RAN 105 and the UPF 115.

[0068] The following will describe Figure 1 Some network functions are shown, which may be involved in some embodiments of the present disclosure.

[0069] In some embodiments, NWDAF 120 may include one or more of the following functions:

[0070] Supports data collection from NF and AF;

[0071] Supports data collection from operations management and maintenance (OAM);

[0072] Register NWDAF services and expose metadata to NF and AF;

[0073] Support the provision of analytical information to NF and AF;

[0074] Supports machine learning (ML) model training and provides ML models (including analysis logic functions) to NWDAF.

[0075] As mentioned above, the maintenance of the FL process between multiple NWDAFs in the 5GC has been added to the conclusions in clause 8.8 of TR 23.700-81 V2.0.0 (mainly in principle 5). The content of principle 5 is as follows:

[0076] Principle 5: The NWDAF, including the MTLF acting as a FL server, can determine the final list of NWDAFs including MTLFs acting as FL clients via the initial FL request to the FL client to determine the availability and compatibility of the FL client. During the FL process, based on local policy or the status of the FL client (e.g., load, availability, capabilities, latency, accuracy, etc.), the NWDAF, including the MTLF acting as a FL server, can trigger the reselection, addition, or removal of FL clients and can announce new FL client discovery via the NRF. FL clients can dynamically join or exit FL operations during the execution phase.

[0077] The solution for maintaining the FL process between multiple NWDAFs in 5GC has been given in TR 23.700-81, 2.0.0 (ie, solution #51).

[0078] Solution #51: Selecting, monitoring, and maintaining NWDAF for federated learning in 5GC

[0079] describe

[0080] This solution is proposed to address key issue #8: supporting federated learning in 5GC. Key research points for this key issue include:

[0081] We study how to coordinate multiple NWDAFs, including selecting the participant NWDAF instances in a federated learning group, e.g., the auxiliary information (if any) used to perform this selection, and the decision on the roles of the participant NWDAFs.

[0082] Study whether and how to perform performance (e.g., network performance and model performance) monitoring of NWDAF federated learning operations.

[0083] To address the challenges of supporting the above-mentioned key points of federated learning in 5GC, this solution focuses on NWDAF selection in the federated learning preparation phase and NWDAF monitoring and maintenance in the federated learning execution phase.

[0084] Many factors influence the selection of client NWDAFs during the federated learning preparation phase, such as NWDAF capabilities, interoperability, and availability of client NWDAFs for joining federated learning.

[0085] During the federated learning execution phase, due to the dynamic changes in the federated network, the current client NWDAF may leave or join. This should account for the dynamic joining and leaving of the client NWDAF in the 5GC during the federated learning / training process. Furthermore, the server NWDAF can also apply methods to monitor the status changes of the client NWDAF (e.g., changes in capabilities and availability).

[0086] process

[0087] Figure 2 is a diagram illustrating an exemplary process for client-side NWDAF selection in the FL preparation phase (where support for FL is applicable) according to some embodiments of the present disclosure.

[0088] In the federated learning preparation phase, the server and (possibly) client NWDAF (e.g. Figure 2 The server NWDAF 120-S and one or more client NWDAFs 120-C-1 to 120-CN (or collectively referred to as 120-C) are shown via NRF (eg, Figure 2 The NRF 160 shown in FIG. 1 is discovered and a client NWDAF is selected by the method used in the handshake mode. The client NWDAF selection is based on availability, capability, etc.

[0089] In some embodiments, an exemplary process for NWDAF selection is Figure 2 is shown and described as follows:

[0090] At step S205, the NWDAF (e.g., the server NWDAF 120-S and / or the client NWDAFs 120-C-1 to 120-CN) may register with the federated learning-enabled NRF 160. In some embodiments, the server NWDAF 120-S may discover the client NWDAF 120-C based on, for example, the federated learning capability, an analysis identifier (ID), and the like.

[0091] At step S210 , the server NWDAF 120 -S may send a federated learning preparation request to the client NWDAF 120 -C, for example by calling the Nnwdaf_MLPreparation_Request service operation with interoperability information. The preparation request may include an indication of the role of the NWDAF (ie, acting as a client NWDAF).

[0092] In some embodiments, the interoperability information may indicate what capabilities the client NWDAF 120-C needs (eg, ability to run certain models) to support the FL process, for example, whether and how the server NWDAF 120-S and the client NWDAF 120-C can share models.

[0093] At step S215 , the client NWDAF 120 -C may decide whether to join the federated learning process based on its availability, capabilities, and interoperability information.

[0094] At step S220 , the client NWDAF 120 -C may send a response to the server NWDAF 120 -S indicating whether it wants to join the FL process.

[0095] At step S225, the server NWDAF 120-S may send a test task to the client NWDAF 120-C that wants to join the FL process. The client NWDAF 120-C may run the test task and send the result to the server NWDAF 120-S.

[0096] In some embodiments, the test task can be a micro-computation or training task, where the requirements for completing the micro-task are the same or similar to those of the main task. In some embodiments, the test task can be a small task that requires the client NWDAF 120-C to collect local data and send local model weights back to the server 120-S; or a test that ensures that the server NWDAF 120-S and the client NWDAF 120-C can communicate if they use the same FL framework or library.

[0097] At step S230 , the server NWDAF 120 -S may select the client NWDAF 120 -C. The server NWDAF 120 -S may select the client NWDAF 120 -C in consideration of the result of the test task.

[0098] Figure 3 is a diagram illustrating an exemplary process for NWDAF monitoring and reselection in the FL execution phase (where support for FL is applicable) according to some embodiments of the present disclosure.

[0099] During the federated learning execution phase, the server NWDAF (server NWDAF 120-S) monitors the client NWDAF (e.g. Figure 3 The status of one or more client NWDAFs 120-C-1 to 120-CN (or collectively referred to as 120-C) is shown to change. In some embodiments, the client NWDAF 120-C can be reselected for FL tasks based on the updated status, availability, and / or capabilities of the client NWDAF 120-C.

[0100] In some embodiments, an exemplary process for monitoring and reselecting a client NWDAF 120-C is described in Figure 3 is shown and described as follows:

[0101] At step S305 , the server NWDAF 120 -S, which monitors the status of the client NWDAF 120 -C during the federated learning execution process, may receive an updated status of the client NWDAF 120 -C.

[0102] In some embodiments, the server NWDAF 120 -S may perform monitoring and obtain the update status of the client NWDAF 120 -C directly and / or via the NRF 160 .

[0103] In some embodiments, the status of the client NWDAF 120-C may be NF load, NF availability, changes in its capabilities (eg, it no longer supports FL).

[0104] At step S310, the server NWDAF 120-S may check the status of the client NWDAF based on the received information and determine whether to reselect the client NWDAF 120-C for the next round of federated learning. In some embodiments, this determination may be based on the updated status of the client NWDAF 120-C, including availability, capabilities, etc.

[0105] At step S315, [if it is determined in step S310 that reselection is required], the server NWDAF 120-S may reselect the client NWDAF 120-C, such as Figure 2 In some embodiments, reference is made to steps S210 to S230. Figure 4 A process for discovering new clients NWDAF 120-C in the federated learning execution phase is described.

[0106] At step S320 , if the client NWDAF 120 -C receives a termination request from the server NWDAF 120 -S, it may terminate the operation for the federated learning.

[0107] There are two possible situations for the server NWDAF to obtain the information of the new client NWDAF, namely: directly obtaining from the new client NWDAF, or obtaining via the NRF.

[0108] In some embodiments, the server NWDAF 120-S may select clients NWDAF #1 120-C-1 to #N 120-CN to participate in the current round of joint learning. In some embodiments, clients NWDAF #N+1 120-C-N+1 to #N+X 120-C-N+X as new clients are capable of joining the next round of training.

[0109] Figure 4FIG2 is a diagram illustrating an exemplary process for dynamically discovering a new NWDAF during the FL execution phase (where FL support is applicable) when information about the server NWDAF is known at the new client NWDAF, according to some embodiments of the present disclosure. In some embodiments, new client NWDAFs (e.g., client NWDAFs #N+1 120-C-N+1 through #N+X 120-C-N+X) that are available and / or capable of joining the federated learning process know information about the server NWDAF 120-S and directly inform the server NWDAF 120-S of the information.

[0110] In some embodiments, the process is as follows Figure 4 Shown and described below:

[0111] At step S405, the server NWDAF 120-S may register the federated learning process with the NRF 160 using the following parameters:

[0112] Federated Learning (FL) related ID.

[0113] Analysis ID.

[0114] In some embodiments, the FL correlation ID can be used to identify a specific FL process. For example, a server NWDAF or a client NWDAF can join different FL processes at the same time, and then when they receive a message or data from another NWDAF, they must know which FL process the message or data is for.

[0115] In some embodiments, when the server NWDAF initiates the FL process, it can register the FL process with the NRF using the FL correlation ID and the analysis ID. When the client NWDAF later wants to dynamically join the FL (for example, it wants to update its local model with global information), it will query the NRF whether there is an ongoing FL for the analysis ID. The NRF will then provide the server NWDAF ID and the FL correlation ID to the client NWDAF, and the client NWDAF can then contact the server NWDAF to join the FL process. Through the FL correlation ID, the server NWDAF knows which FL process the client NWDAF wants to join and which model should be provided to the client.

[0116] At step S410, if information about the server NWDAF 120-S and the corresponding FL process is known via the NRF 160, the new clients NWDAFs 120-C-N+1 to 120-C-N+X may inform the server NWDAF 120-S of their interoperability and availability by calling the Nnwdaf_MLPreparation_Request service operation.

[0117] At step S415, before starting the next round of training, the server NWDAF 120-S may select a client NWDAF from NWDAF #1 120-C-1 to #N+X 120-C-N+X based on the updated information of the client NWDAF 120-C. In some embodiments, this process may be similar to Figure 2 Steps S210 to S230 are the same as in FIG.

[0118] Figure 5 This diagram illustrates an exemplary process for dynamically discovering a new NWDAF during the FL execution phase (where FL support is applicable) when information about the server NWDAF is unknown at the new client NWDAF, according to some embodiments of the present disclosure. In some embodiments, the server NWDAF 120-S may select client NWDAFs #1 120-C-1 through #N 120-CN to participate in the current round of federated learning. In some embodiments, client NWDAFs #N+1 120-C-N+1 through #N+X 120-C-N+X, as new clients, may be able to join the next round of training.

[0119] In some embodiments, the process is as follows Figure 5 Shown and described below:

[0120] Similar to step S405, at step S505, the server NWDAF 120-S may register the federated learning process with the NRF 160 using the following parameters:

[0121] Federated Learning (FL) related ID.

[0122] Analysis ID.

[0123] At step S510 , the server NWDAF 120 -S may obtain new client NWDAF information by subscribing to an event of registration of a new client NWDAF or discovering the client NWDAF via the NRF 160 .

[0124] Similar to step S415, at step S515, before starting the next round of training, the server NWDAF 120-S may select a client NWDAF from NWDAF #1 120-C-1 to #N+X 120-C-N+X based on the updated information of the client NWDAF 120-C. In some embodiments, this process may be similar to Figure 2 Steps S210 to S230 are the same as in FIG.

[0125] In some embodiments, the server NWDAF 120-S can dynamically obtain information about the new client NWDAF via the NRF 160 by subscribing to an event indicating registration of a new client NWDAF or discovering the NRF 160 when reselecting a client NWDAF is required. However, Solution #51 does not provide details on how the server NWDAF 120-S obtains analytical data (e.g., NF load, etc.) of the client NWDAF 120-C or how to terminate the federated learning operation at the client NWDAF 120-C.

[0126] As mentioned above, the maintenance of the FL process between multiple NWDAFs in the 5GC has been added to the conclusions in clause 8.8 of TR 23.700-81 (primarily in principle 5).

[0127] Principle 5: The NWDAF, including the MTLF acting as a FL server, can determine the final list of NWDAFs including MTLFs acting as FL clients via the initial FL request to the FL client to determine the availability and compatibility of the FL client. During the FL process, based on local policy or the status of the FL client (e.g., load, availability, capabilities, latency, accuracy, etc.), the NWDAF, including the MTLF acting as a FL server, can trigger the reselection, addition, or removal of FL clients and can announce new FL client discovery via the NRF. FL clients can dynamically join or exit FL operations during the execution phase.

[0128] However, for the maintenance and implementation of the FL process, it remains unclear how the server NWDAF obtains the analytical data (e.g., NF load, etc.) of the client NWDAF and how to terminate the federated learning operation at the client NWDAF.

[0129] In some embodiments of the present disclosure, maintenance and implementation of a complete FL process are proposed. In some embodiments, a process is provided for a server NWDAF to obtain analysis data of a client NWDAF and terminate a federated learning operation at the client NWDAF.

[0130] In some embodiments, when the server NWDAF obtains the analysis data (for example, NF load, etc.) of the client NWDAF, the following situations may be considered:

[0131] The server NWDAF subscribes to some other NWDAFs (referred to as “auxiliary NWDAFs”) for the analysis data of the client NWDAF. In some embodiments, the auxiliary NWDAFs may notify the server NWDAF of the analysis results.

[0132] The server NWDAF subscribes to the analysis data from the client NWDAF. The client NWDAF can perform analysis on its own and notify the server NWDAF of the analysis results.

[0133] In some embodiments, the analysis data may be sent to the server NWDAF periodically or dynamically in a notification when a certain predetermined state is reached.

[0134] In some embodiments, two possible scenarios may be considered for terminating the federated learning operation at the client NWDAF:

[0135] The client NWDAF exits the joint learning process;

[0136] The server NWDAF removes the client NWDAF from the federated learning process.

[0137] In some embodiments, for the above two processes, two methods of exchanging ML model information between the server NWDAF and the client NWDAF in the FL execution phase can be considered respectively:

[0138] Reuse existing services (or their extensions), such as Nnwdaf_MLModelProvision given in TS 23.288 V18.0.0, for ML model information exchange between server NWDAF and client NWDAF.

[0139] Use the new service to exchange ML model information between server NWDAF and client NWDAF.

[0140] In some embodiments, maintenance and implementation of a complete federated learning process are proposed. In some embodiments, processes for the server NWDAF to obtain client NWDAF analysis data through an auxiliary NWDAF and directly from the client NWDAF are presented. In some embodiments, processes for terminating the federated learning operation at the client NWDAF are presented, while considering two different methods for exchanging ML model information during the FL execution phase. In some embodiments, the two methods for exchanging ML model information include:

[0141] Reuse existing services (or their extensions), such as Nnwdaf_MLModelProvision, for ML model information exchange between server NWDAF and client NWDAF.

[0142] In some embodiments, a new service is used to exchange ML model information between the server NWDAF and the client NWDAF.

[0143] The maintenance of the federated learning process has been concluded in TR 23.700-81, but some unclear points remain, such as how the server NWDAF obtains the analysis data of the client NWDAF and how to terminate the federated learning operation at the client NWDAF. Through some embodiments of the present disclosure, the maintenance and implementation of the federated learning process are proposed. The process of the server NWDAF obtaining the client NWDAF analysis data through the auxiliary NWDAF and the process of obtaining the analysis data directly from the client NWDAF are respectively given. Through some embodiments of the present disclosure, the process of terminating the federated learning operation at the client NWDAF is given, and two different methods for exchanging ML model information in the FL execution phase are considered, namely, reusing existing services (or their extensions) and using new services.

[0144] Figure 6 is a diagram illustrating an exemplary system for analyzing data collection according to some embodiments of the present disclosure. Figure 6 An exemplary system is shown for a server NWDAF 120-S to obtain analytics data of a client NWDAF 120-C via an auxiliary NWDAF 120-A.

[0145] like Figure 6 As shown, the server NWDAF 120-S may subscribe to the auxiliary NWDAF 120-A for analysis data about the client NWDAF 120-C, such as NF load, etc. The auxiliary NWDAF 120-A may perform analysis on the client NWDAF 120-C and dynamically notify the server NWDAF 120-S of the analysis periodically or when a predetermined state is reached.

[0146] Alternatively, the server NWDAF 120-S may subscribe to the client NWDAF 120-C for analysis data, and the client NWDAF 120-C may perform analysis on itself and dynamically notify the server NWDAF 120-S of the analysis periodically or upon reaching a certain predetermined state.

[0147] In some embodiments, the FL process may be ongoing between the server NWDAF 120-S and the client NWDAF 120-C, or the client NWDAF 120-C may be a new candidate client NWDAF that the server NWDAF 120-S selects for the FL process.

[0148] Figure 7 is a diagram illustrating an exemplary system for terminating the FL process at a client NWDAF according to some embodiments of the present disclosure. Figure 7An exemplary system for terminating a federated learning operation at the client NWDAF 120 -C in the FL execution phase is shown.

[0149] exist Figure 7 In (a), the client NWDAF 120-C may exit the federated learning operation during the FL execution phase. Figure 7 In (b), the server NWDAF 120 -S may terminate the federated learning operation at the client NWDAF 120 -C to be removed from the FL process.

[0150] Figure 8 is a diagram illustrating an exemplary scenario for analyzing data collection according to some embodiments of the present disclosure. Figure 8 Exemplary systems are shown for a server NWDAF 120-S to obtain analytics data of a client NWDAF 120-C through an auxiliary NWDAF 120-A (eg, as shown in (a)) and directly from the client NWDAF 120-C (eg, as shown in (b)).

[0151] like Figure 8 As shown in (a) of FIG5 , the server NWDAF 120-S may subscribe to analysis data about the client NWDAF 120-C from the auxiliary NWDAF 120-A, for example, analysis ID = "NF load information". The auxiliary NWDAF 120-A may collect data from other NFs (which may also be the client NWDAF 120-C), AFs, and / or OAMs 810 in the 5GC, and perform analysis on the analysis ID of the client NWDAF 120-C based on the collected data, and periodically or dynamically (for example, when a predetermined state is reached) notify the server NWDAF 120-S of the analysis results.

[0152] like Figure 8 As shown in (b), the server NWDAF 120-S may subscribe to self-analysis from the client NWDAF 120-C, for example, analysis ID = "NF load information". The client NWDAF 120-C may collect data from other NFs, AFs, and / or OAM 810 in the 5GC, and perform analysis for its own analysis ID based on the collected data and possibly also based on its own data, and notify the server NWDAF 120-S of the analysis results periodically or dynamically (for example, when a predetermined state is reached).

[0153] In some embodiments, the FL process may be ongoing between the server NWDAF 120-S and the client NWDAF 120-C, or the client NWDAF 120-C may be a new candidate client NWDAF that the server NWDAF 120-S selects for the FL process.

[0154] Figure 9 is a diagram illustrating an exemplary scenario for terminating the FL process at the client NWDAF according to some embodiments of the present disclosure. Figure 9 An exemplary system is shown for terminating a federated learning operation at a client NWDAF 120-C in an FL execution phase by reusing an existing service (or an extension thereof) as shown in (a), exiting the client NWDAF 120-C using the new service as shown in (b), and terminating the server NWDAF 120-S using the new service as shown in (c).

[0155] like Figure 9 As shown in (a) of FIGURE 2, server NWDAF 120-S and client NWDAF 120-C can unsubscribe from each other's ML model information exchange during the FL execution phase by calling the Nnwdaf_MLModelProvision_Unsubscribe service operation. Upon receiving the unsubscribe request, server NWDAF 120-S can stop sharing ML model information with client NWDAF 120-C. Client NWDAF 120-C can terminate the corresponding federated learning operation and stop sharing local ML model information with server NWDAF 120-S.

[0156] like Figure 9 As shown in (b) of FIG1 , the client NWDAF 120-C may send an Nnwdaf_MLTraining_Quit request to the server NWDAF 120-S to exit the federated learning process. Upon receiving the exit request from the client NWDAF 120-C, the server NWDAF 120-S may stop sharing ML model information with the client NWDAF 120-C and send an Nnwdaf_MLTraining_Quit response to the client NWDAF 120-C. Upon receiving the exit response, the client NWDAF 120-C may terminate the corresponding federated learning operation and stop sharing local ML model information with the server NWDAF 120-S.

[0157] like Figure 9As shown in (c) of FIG5 , the server NWDAF 120-S may send a Nnwdaf_MLTraining_Terminate request to the client NWDAF 120-C to terminate the federated learning operation at the client NWDAF 120-C and stop sharing the ML model information with the client NWDAF 120-C. After receiving the termination request from the server NWDAF 120-S, the client NWDAF 120-C may terminate the corresponding federated learning operation, stop sharing the local ML model information with the server NWDAF 120-S, and send a Nnwdaf_MLTraining_Terminate response to the server NWDAF 120-S.

[0158] Figure 10 is a diagram illustrating an exemplary process for analyzing data collection according to some embodiments of the present disclosure. Figure 10 Two exemplary processes for the server NWDAF 120-S to obtain analysis data of the client NWDAF 120-C in the federated learning execution phase are shown, corresponding to the following two cases:

[0159] Case 1: The server NWDAF 120-S obtains the analysis data of the client NWDAF 120-C from the auxiliary NWDAF 120-A;

[0160] Case 2: The server NWDAF 120-S obtains the analysis data of the client NWDAF 120-C from the client NWDAF 120-C.

[0161] The corresponding process is described as follows.

[0162] Case 1:

[0163] At step S1005, the server NWDAF 120-S may subscribe the analytics data of the client NWDAF 120-C to the secondary NWDAF 120-A by calling the Nnwdaf_AnalyticsSubscription_Subscribe service operation (analysis ID = "NF load information", etc.). In some embodiments, clauses 6.1.1 and 7.2.2 of TS 23.288 V18.0.0 (regarding Nnwdaf_AnalyticsSubscription_Subscribe) provide an exemplary procedure for establishing such a subscription.

[0164] At step S1010, the auxiliary NWDAF 120-A may collect data for analysis from other NFs, AFs, OAM 810, etc., and may also collect data from the client NWDAF 120-C. In some embodiments, clauses 6.3 to 6.16 of TS 23.288 V18.0.0 (regarding analysis ID and input) provide exemplary information on data collection for analysis.

[0165] At step S1015 , after data collection, the auxiliary NWDAF 120 -A may perform analysis on the analysis ID (eg, NF load information) of the client NWDAF 120 -C based on the collected data.

[0166] At step S1020, the auxiliary NWDAF 120-A may notify the server NWDAF 120-S of the analysis results by periodically or dynamically (e.g., upon reaching a predetermined state) invoking the Nnwdaf_AnalyticsSubscription_Notify service operation. In some embodiments, clauses 6.1.1 and 7.2.4 of TS 23.288 V18.0.0 (regarding Nnwdaf_AnalyticsSubscription_Notify) provide exemplary procedures for sending notifications. In some embodiments, clauses 6.3 through 6.16 of TS 23.288 V18.0.0 (regarding analysis ID and output) provide exemplary information regarding outputs.

[0167] Case 2:

[0168] At step S1025, the server NWDAF 120-S may subscribe the analytics data of the client NWDAF 120-C to (all or some of) the client NWDAFs 120-C by calling the Nnwdaf_AnalyticsSubscription_Subscribe service operation (analysis ID = "NF load information", etc.). In some embodiments, an exemplary procedure for establishing this subscription is provided in clauses 6.1.1 and 7.2.2 of TS 23.288 V18.0.0 (regarding Nnwdaf_AnalyticsSubscription_Subscribe).

[0169] At step S1030 , the client NWDAF 120 -C may collect data for analysis from other NFs, AFs, OAM 810 , etc. In some embodiments, clauses 6.3 to 6.16 of TS 23.288 V18.0.0 (regarding analysis ID and input) provide exemplary information on data collection for analysis.

[0170] At step S1035 , after data collection, the client NWDAF 120 -C may perform analysis on the analysis ID (eg, NF load information) based on the collected data by itself, and may also perform analysis on its own data.

[0171] At step S1040, the client NWDAF 120-C may notify the server NWDAF 120-S of the analysis results by periodically or dynamically (e.g., upon reaching a predetermined state) invoking the Nnwdaf_AnalyticsSubscription_Notify service operation. In some embodiments, clauses 6.1.1 and 7.2.4 of TS 23.288 V18.0.0 (regarding Nnwdaf_AnalyticsSubscription_Notify) provide exemplary procedures for sending notifications. In some embodiments, clauses 6.3 through 6.16 of TS 23.288 V18.0.0 (regarding analysis ID and output) provide exemplary information regarding outputs.

[0172] Figure 11A and Figure 11B is a diagram illustrating an exemplary process for terminating the FL process at the client NWDAF according to some embodiments of the present disclosure. Figure 11A and Figure 11B A process for terminating a federated learning operation at the client NWDAF 120-C in the FL execution phase is shown, corresponding to the following four cases:

[0173] Case 1: Using an existing service (or its extension) for ML model information exchange, the client(s) NWDAF 120 - C decide to exit the FL process;

[0174] Case 2: Using existing services (or their extensions) for ML model information exchange, the server NWDAF 120 -S decides to remove the client(s) NWDAF 120 -C from the FL process;

[0175] Case 3: Using a new service for ML model information exchange, the client(s) NWDAF 120-C decides to exit the FL process;

[0176] Case 4: Using a new service for ML model information exchange, the server NWDAF 120 -S decides to remove the client(s) NWDAF 120 -C from the FL process.

[0177] The corresponding process is described as follows.

[0178] like Figure 11A and Figure 11BAs shown, at step S1105 , the federated learning process is ongoing, and ML model information is exchanged between the server NWDAF 120 -S and the client NWDAF 120 -C in the FL execution phase by using an existing service (or an extension thereof) (e.g., Nnwdaf_MLModelProvision service) or a new service (e.g., Nnwdaf_MLTraining service or any other possible new service).

[0179] The remaining steps for cases 1 to 4 are described below.

[0180] Figure 11A in Case 1 :

[0181] At step S1110 , the client NWDAF 120 -C may decide to exit the FL process.

[0182] At step S1115 , the client NWDAF 120 -C may unsubscribe from the ML model information exchange to the server NWDAF 120 -S by calling the Nnwdaf_MLModelProvision_Unsubscribe service operation using the FL correlation ID and a reason code (e.g., the client NWDAF 120 -C exits the FL process and specifies a detailed reason (e.g., availability change, capability change, etc.)).

[0183] In some embodiments, the client NWDAF 120-C may send a Nnwdaf_MLModelProvision_Unsubscribe request message to the server NWDAF 120-S, and the server NWDAF 120-S may respond with a Nnwdaf_MLModelProvision_Unsubscribe response message. In other embodiments, the client NWDAF 120-C may send a Nnwdaf_MLModelProvision_Unsubscribe request message to the server NWDAF 120-S, and the server NWDAF 120-S may respond with another Nnwdaf_MLModelProvision_Unsubscribe request message instead of a Nnwdaf_MLModelProvision_Unsubscribe response message. In either case, the client and server NWDAFs may agree to terminate the FL process at the client NWDAF.

[0184] In some embodiments, if the client NWDAF 120-C exits, the context of the FL process at the client NWDAF 120-C will be cleared.

[0185] In some embodiments, clauses 6.2A.1 and 7.5.3 of TS 23.288 V18.0.0 (regarding Nnwdaf_MLModelProvision_Unsubscribe) provide an exemplary description of unsubscribing.

[0186] At step S1120 , the server NWDAF 120 -S and the client NWDAF 120 -C may stop FL operations with respect to each other.

[0187] At step 1120a, the server NWDAF 120-S may stop the federated learning operation for the FL process with respect to the client NWDAF 120-C.

[0188] At step 1120b, the client NWDAF 120-C may stop the federated learning operations related to the FL process.

[0189] Figure 11A in Case 2 :

[0190] At step S1125 , the server NWDAF 120 -S may decide to remove the client NWDAF 120 -C from the FL process.

[0191] In some embodiments, the client NWDAF 120 -C may also decide to exit the FL process.

[0192] At step S1130 , the server NWDAF 120 -S may unsubscribe the client NWDAF 120 -C from the ML model information exchange by calling the Nnwdaf_MLModelProvision_Unsubscribe service operation using the FL correlation ID and a reason code (e.g., the client NWDAF 120 -C was deselected by the server NWDAF 120 -S for the FL process, or the FL process is suspended, etc.).

[0193] In some embodiments, the server NWDAF 120-S may send a Nnwdaf_MLModelProvision_Unsubscribe request message to the client NWDAF 120-C, and the client NWDAF 120-C may respond with a Nnwdaf_MLModelProvision_Unsubscribe response message. In some other embodiments, the server NWDAF 120-S may send a Nnwdaf_MLModelProvision_Unsubscribe request message to the client NWDAF 120-C, and the server NWDAF 120-S may respond with another Nnwdaf_MLModelProvision_Unsubscribe request message instead of a Nnwdaf_MLModelProvision_Unsubscribe response message. In either case, the client and server NWDAFs may agree to terminate the FL process at the client NWDAF.

[0194] In some embodiments, if the client NWDAF 120 -C is deselected during a FL process, the context of the FL process at the client NWDAF 120 -C will be cleared.

[0195] In some embodiments, an exemplary description of unsubscribing is provided in clauses 6.2A.1 and 7.5.3 of TS 23.288 V18.0.0 (regarding Nnwdaf_MLModelProvision_Unsubscribe).

[0196] At step S1135 , the server NWDAF 120 -S and the client NWDAF 120 -C may stop FL operations with respect to each other.

[0197] At step S1135a, the server NWDAF 120-S may stop the federated learning operation for the FL process with respect to the client NWDAF 120-C.

[0198] At step S1135b, the client NWDAF 120-C may stop the federated learning operation related to the FL process.

[0199] Figure 11B in Case 3 :

[0200] At step S1140 , the client NWDAF 120 -C may decide to exit the federated learning process.

[0201] At step S1145 , the client NWDAF 120 -C may send a request to exit the FL process to the server NWDAF 120 -S by calling, for example, the Nnwdaf_MLTraining_Quit request service operation (or using any other possible new service for ML model information exchange in the FL execution phase) using the FL correlation ID and a reason code (e.g., availability change, capability change, etc.).

[0202] At step S1150 , the server NWDAF 120 -S may decide to remove the client NWDAF 120 -C from the FL process, stopping the FL operation related to the client NWDAF 120 -C.

[0203] At step S1155 , the server NWDAF 120 -S may respond to the quit request to the client NWDAF 120 -C by calling the Nnwdaf_MLTraining_Quit response service operation (or using any other possible new service for ML model information exchange in the FL execution phase) with parameters (e.g., the time when the client NWDAF 120 -C quits, etc.).

[0204] At step S1160 , the client NWDAF 120 -C may stop the federated learning operation and clear the context related to the FL process.

[0205] Figure 11B in Case 4 :

[0206] At step S1165 , the server NWDAF 120 -S may decide to remove the client NWDAF 120 -C from the FL process.

[0207] At step S1170 , the server NWDAF 120 -S may send a request for a FL operation to terminate the FL process at the client NWDAF 120 -C by calling, for example, the Nnwdaf_MLTraining_Terminate request service operation (or using any other possible new service for ML model information exchange in the FL execution phase) using the FL correlation ID and a reason code (e.g., the client NWDAF 120 -C was deselected for the FL process by the server NWDAF 120 -S, or the FL process is suspended, etc.).

[0208] At step S1175 , the client NWDAF 120 -C may stop the federated learning operation related to the FL process.

[0209] In some embodiments, if the client NWDAF 120 -C is deselected from the FL process, the context of the FL process at the client NWDAF 120 -C will be cleared.

[0210] At step S1180 , the client NWDAF 120 -C may respond to the termination request to the server NWDAF 120 -S by calling the Nnwdaf_MLTraining_Terminate response service operation (or using any other possible new service for ML model information exchange in the FL execution phase).

[0211] At step S1185 , the server NWDAF 120 -S may stop the federated learning operation for the FL process with respect to the client NWDAF 120 -C.

[0212] According to some of the aforementioned embodiments of the present disclosure, federated learning can be supported in core networks such as 5GC networks. This approach enables the maintenance and implementation of a federated learning process. Separate procedures are provided for the server NWDAF to obtain client NWDAF analysis data via an auxiliary NWDAF and for obtaining analysis data directly from the client NWDAF. Furthermore, procedures are provided for terminating the federated learning operation at the client NWDAF. Two different approaches for exchanging ML model information during the FL execution phase are considered: reusing existing services (or extensions thereof) and using new services.

[0213] Figure 12 is a flow chart of an example method 1200 at a server associated with an FL process according to an embodiment of the present disclosure. The method 1200 may be performed at a server NWDAF (e.g., Figure 6 The method 1200 may include at least one of steps S1210 and S1220. However, the present disclosure is not limited thereto. In some other embodiments, the method 1200 may include more steps, fewer steps, different steps, or any combination thereof. Furthermore, the steps of the method 1200 may be performed in an order different from that described herein. Furthermore, in some embodiments, the steps of the method 1200 may be broken down into multiple sub-steps and performed by different entities, and / or multiple steps of the method 1200 may be combined into a single step.

[0214] Method 1200 may start with at least one of steps S1210 and S1220.

[0215] At step S1210 , the server may send a first message to one or more clients associated with the FL process, the first message indicating that the corresponding client is deselected by the server for the FL process and / or the FL process is suspended.

[0216] At step S1220 , the server may receive a second message from one or more clients associated with the FL process, the second message indicating that the corresponding client is exiting the FL process.

[0217] In some embodiments, at least one of the first message and the second message may further indicate at least one of the following: a FL correlation ID; and a reason code. In some embodiments, the reason code may indicate at least one of the following: when the reason code is indicated by the first message, the corresponding client is deselected by the server for the FL process; when the reason code is indicated by the first message, the FL process is suspended; when the reason code is indicated by the corresponding second message, an availability change associated with the corresponding client; and when the reason code is indicated by the corresponding second message, a capability change associated with the corresponding client.

[0218] In some embodiments, method 1200 may further include at least one of the following: receiving a third message from at least one of the one or more clients in response to the corresponding first message, the third message indicating that the FL process has terminated or is to be terminated at the at least one client; and sending a fourth message to at least one of the one or more clients in response to the corresponding second message, the fourth message indicating that the at least one client has been removed or is to be removed from the FL process. In some embodiments, the fourth message may further indicate a time when the at least one client exits the FL process.

[0219] In some embodiments, method 1200 may further include at least one of the following: in response to sending a corresponding first message, stopping one or more FL operations associated with at least one of the one or more clients of the FL process; in response to receiving a corresponding second message, stopping one or more FL operations associated with at least one of the one or more clients of the FL process; and in response to receiving a corresponding third message, stopping one or more FL operations associated with at least one of the one or more clients of the FL process. In some embodiments, the first message may cause the client to stop the one or more FL operations of the FL process. In some embodiments, at least one of the first message, the second message, the third message, and the fourth message may be a message defined in 3GPP TS 23.288, V18.0.0, and / or any previous version thereof, or an extension thereof. In some embodiments, at least one of the first message, the second message, the third message, and the fourth message may not be a message defined in 3GPP TS 23.288 V18.0.0, and / or any previous version thereof, or an extension thereof.

[0220] In some embodiments, the first message may be one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Terminate request message, and / or the third message may be a corresponding message among a Nnwdaf_MLModelProvision_Unsubscribe request message, a Nnwdaf_MLModelProvision_Unsubscribe response message, and a Nnwdaf_MLTraining_Terminate response message. In some embodiments, the second message may be one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Quit request message, and / or the fourth message may be a corresponding message among a Nnwdaf_MLModelProvision_Unsubscribe request message, a Nnwdaf_MLModelProvision_Unsubscribe response message, and a Nnwdaf_MLTraining_Quit response message.

[0221] In some embodiments, before the step of sending the first message and / or the step of receiving the second message, method 1200 may further include at least one of the following: sending ML model information to the one or more clients; and receiving ML model information from the one or more clients. In some embodiments, the NWDAF may be hosted by a server. In some embodiments, one or more NWDAFs may be hosted by the one or more clients. In some embodiments, method 1200 may further include referring to Figure 13 Any step in any method described.

[0222] Figure 13 is a flow chart of an example method 1300 at a server associated with an FL process according to an embodiment of the present disclosure. The method 1300 may be performed at a server NWDAF (e.g., Figure 6 The method 1300 may include steps S1310 and S1320. However, the present disclosure is not limited thereto. In some other embodiments, the method 1300 may include more steps, different steps, or any combination thereof. Furthermore, the steps of the method 1300 may be performed in an order different from that described herein. Furthermore, in some embodiments, the steps of the method 1300 may be broken down into multiple sub-steps and performed by different entities, and / or multiple steps of the method 1300 may be combined into a single step.

[0223] Method 1300 may begin at step S1310 , where the server may receive one or more fifth messages indicating analysis data associated with one or more clients in the FL process and / or one or more candidate clients to be selected for the FL process.

[0224] At step S1320 , the server may select at least one client from the one or more candidate clients and / or the one or more clients for the FL process based at least on the analysis data.

[0225] In some embodiments, the one or more fifth messages may be received from at least one of: at least one of the one or more clients; at least one of the one or more candidate clients; and one or more network nodes. In some embodiments, the one or more network nodes may be network nodes that assist the server in collecting analytical data associated with at least one of the one or more clients and / or at least one of the one or more candidate clients.

[0226] In some embodiments, before receiving the one or more fifth messages, method 1300 may further include at least one of the following: sending a sixth message to at least one of the one or more clients and / or at least one of the one or more candidate clients, the sixth message being used to subscribe to analytics data associated with the at least one client and / or at least one of the one or more candidate clients; and sending a sixth message to at least one of the one or more network nodes, the sixth message being used to subscribe to analytics data associated with the at least one of the one or more clients and / or at least one of the one or more candidate clients. In some embodiments, the fifth message may be a Nnwdaf_AnalyticsSubscription_Notify request message, and the sixth message may be a Nnwdaf_AnalyticsSubscription_Subscribe request message.

[0227] In some embodiments, the analysis data may include data related to at least one of the following: load, availability, capacity, latency, and accuracy. In some embodiments, the step of receiving the one or more fifth messages may be performed periodically and / or dynamically in response to an event. In some embodiments, the NWDAF may be hosted by a server. In some embodiments, the NWDAF may be hosted by the one or more clients. In some embodiments, the NWDAF may be hosted by the one or more candidate clients. In some embodiments, the method 1300 may further include referencing Figure 12Any step in any method described.

[0228] Figure 14 is a flow chart of an example method 1400 at a client associated with an FL process according to an embodiment of the present disclosure. The method 1400 may be performed at a client NWDAF (e.g., Figure 6 Method 1400 may include at least one of steps S1410 and S1420. However, the present disclosure is not limited thereto. In some other embodiments, method 1400 may include more steps, fewer steps, different steps, or any combination thereof. Furthermore, the steps of method 1400 may be performed in an order different from that described herein. Furthermore, in some embodiments, the steps of method 1400 may be broken down into multiple sub-steps and performed by different entities, and / or multiple steps of method 1400 may be combined into a single step.

[0229] Method 1400 may start with at least one of steps S1410 and S1420.

[0230] At step S1410 , the client may receive a first message from a server associated with the FL process, the first message indicating that the client is deselected by the server for the FL process and / or the FL process is suspended.

[0231] At step S1420 , the client may send a second message to the server associated with the FL process, the second message indicating that the client is exiting the FL process.

[0232] In some embodiments, at least one of the first message and the second message may further indicate at least one of the following: a FL correlation ID; and a reason code. In some embodiments, the reason code may indicate at least one of the following: when the reason code is indicated by the first message, the client is deselected by the server for the FL process; when the reason code is indicated by the first message, the FL process is suspended; when the reason code is indicated by the corresponding second message, an availability change associated with the client; and when the reason code is indicated by the second message, a capability change associated with the client.

[0233] In some embodiments, method 1400 may further include at least one of the following: sending a third message to the server in response to the first message, the third message indicating that the FL process has terminated or is about to terminate at the client; and receiving a fourth message from the server in response to the second message, the fourth message indicating that the client has been removed or is about to be removed from the FL process. In some embodiments, the fourth message may further indicate the time when the client exited the FL process. In some embodiments, method 1400 may further include at least one of the following: stopping one or more FL operations of the FL process in response to receiving the first message; clearing the context of the FL process in response to receiving the first message; stopping one or more FL operations of the FL process in response to sending the second message; clearing the context of the FL process in response to sending the second message; and stopping one or more FL operations of the FL process in response to receiving the fourth message; clearing the context of the FL process in response to receiving the fourth message.

[0234] In some embodiments, at least one of the first message, the second message, the third message, and the fourth message may be a message defined in 3GPP TS 23.288, V18.0.0, and / or any previous version thereof, or an extension of the message. In some embodiments, at least one of the first message, the second message, the third message, and the fourth message may be neither a message defined in 3GPP TS 23.288 V18.0.0, and / or any previous version thereof, nor an extension of the message.

[0235] In some embodiments, the first message may be one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Terminate request message, and / or the third message may be a corresponding message among a Nnwdaf_MLModelProvision_Unsubscribe request message, a Nnwdaf_MLModelProvision_Unsubscribe response message, and a Nnwdaf_MLTraining_Terminate response message. In some embodiments, the second message may be one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Quit request message, and / or the fourth message may be a corresponding message among a Nnwdaf_MLModelProvision_Unsubscribe request message, a Nnwdaf_MLModelProvision_Unsubscribe response message, and a Nnwdaf_MLTraining_Quit response message.

[0236] In some embodiments, before the step of receiving the first message and / or the step of sending the second message, method 1400 may further include at least one of the following: receiving ML model information from the server; and sending ML model information to the server. In some embodiments, NWDAF may be hosted by the client. In some embodiments, NWDAF is hosted by the server. In some embodiments, method 1400 may further include referring to Figure 15 Any step in any method described.

[0237] Figure 15 is a flow chart of an example method 1500 at a client associated with a FL process or at a candidate client to be selected for a FL process according to an embodiment of the present disclosure. The method 1500 may be performed at a client NWDAF (e.g., Figure 6 Method 1500 may include at least one of steps S1510 and S1520. However, the present disclosure is not limited thereto. In some other embodiments, method 1500 may include more steps, fewer steps, different steps, or any combination thereof. Furthermore, the steps of method 1500 may be performed in an order different from that described herein. Furthermore, in some embodiments, the steps of method 1500 may be broken down into multiple sub-steps and performed by different entities, and / or multiple steps of method 1500 may be combined into a single step.

[0238] Method 1500 may start with at least one of steps S1510 and S1520.

[0239] At step S1510 , the client or candidate client may send a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with the client or candidate client.

[0240] At step S1520 , the client or candidate client may send a seventh message to one or more network nodes, the seventh message indicating data associated with the client or candidate client, which may be used as input data when determining analysis data associated with the client or candidate client.

[0241] In some embodiments, the one or more network nodes may be network nodes that assist the server in collecting analytics data associated with the client or candidate client. In some embodiments, before sending the fifth message, method 1500 may further include at least one of the following: receiving a sixth message from the server, the sixth message subscribing to analytics data associated with the client or candidate client; and receiving an eighth message from at least one of the one or more network nodes, the eighth message requesting input data associated with the client or candidate client. In some embodiments, method 1500 may further include collecting data for determining analytics data associated with the client or candidate client from at least one of one or more network functions (NFs), one or more application servers (AFs), and one or more OAM nodes; and performing analysis on the collected data to determine analytics data associated with the client or candidate client.

[0242] In some embodiments, the fifth message may be a Nnwdaf_AnalyticsSubscription_Notify request message, wherein the sixth message may be a Nnwdaf_AnalyticsSubscription_Subscribe request message. In some embodiments, the analytical data may include data related to at least one of the following items: load, availability, capacity, latency, and accuracy. In some embodiments, the step of sending the fifth message may be performed periodically and / or dynamically in response to an event. In some embodiments, the NWDAF may be hosted by the client or candidate client. In some embodiments, the NWDAF may be hosted by a server. In some embodiments, the method 1500 may also include a reference to Figure 14 Any step in any method described.

[0243] Figure 16 is a flow chart of an example method 1600 at a network node according to an embodiment of the present disclosure. The method 1600 may be performed in a secondary NWDAF (e.g., Figure 6 Method 1600 may include step S1610. However, the present disclosure is not limited thereto. In some other embodiments, method 1600 may include more steps, different steps, or any combination thereof. Furthermore, the steps of method 1600 may be performed in an order different from that described herein. Furthermore, in some embodiments, the steps of method 1600 may be split into multiple sub-steps and performed by different entities, and / or multiple steps of method 1600 may be combined into a single step.

[0244] Method 1600 may begin at step S1610, where the network node may send a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with: one or more clients associated with the FL process and / or one or more candidate clients to be selected by the server for the FL process.

[0245] In some embodiments, the network node may be a network node that assists the server in collecting analysis data associated with the one or more clients and / or the one or more candidate clients. In some embodiments, before sending the fifth message, method 1600 may further include: receiving a sixth message from the server, the sixth message being used to subscribe to analysis data associated with the one or more clients and / or the one or more candidate clients.

[0246] In some embodiments, the fifth message may be a Nnwdaf_AnalyticsSubscription_Notify request message, wherein the sixth message may be a Nnwdaf_AnalyticsSubscription_Subscribe request message. In some embodiments, before sending the fifth message, method 1600 may further include: collecting data for determining analytical data associated with the one or more clients and / or the one or more candidate clients from at least one of the one or more clients, the one or more candidate clients, the one or more NFs, the one or more AFs, and the one or more OAM nodes; and performing analysis on the collected data to determine analytical data associated with the one or more clients and / or the one or more candidate clients.

[0247] In some embodiments, the analysis data may include data related to at least one of the following: load, availability, capacity, latency, and accuracy. In some embodiments, the step of sending the fifth message may be performed periodically and / or dynamically in response to an event. In some embodiments, the NWDAF may be hosted by the client and / or candidate client. In some embodiments, the NWDAF may be hosted by a server.

[0248] Figure 17Schematically illustrates an embodiment of an arrangement that can be used in a server, client, and / or network node according to an embodiment of the present disclosure. Arrangement 1700 includes a processing unit 1706, for example, having a digital signal processor (DSP) or a central processing unit (CPU). Processing unit 1706 can be a single unit or multiple units for performing the different actions of the processes described herein. Arrangement 1700 can also include an input unit 1702 for receiving signals from other entities and an output unit 1704 for providing signals to other entities. Input unit 1702 and output unit 1704 can be arranged as an integrated entity or as separate entities.

[0249] Furthermore, the arrangement 1700 may include at least one computer program product 1708 in the form of a non-volatile or volatile memory, such as an electrically erasable programmable read-only memory (EEPROM), a flash memory and / or a hard drive. The computer program product 1708 includes a computer program 1710 comprising code / computer-readable instructions which, when executed by the processing unit 1706 in the arrangement 1700, causes the arrangement 1700 and / or the servers and / or clients and / or network nodes included therein to perform, for example, the operations previously described in connection with Figures 2 to 16 or any other variation describing the actions of the process.

[0250] The computer program 1710 may be configured as computer program code implemented in computer program modules 1710A and / or 1710B. Thus, in an exemplary embodiment, when the arrangement 1700 is used in a server associated with a FL process, the code in the computer program of the arrangement 1700 includes at least one of the following: a module 1710A configured to send a first message to one or more clients associated with the FL process, the first message indicating that the corresponding client has been deselected by the server for the FL process and / or that the FL process is suspended; and a module 1710B configured to receive a second message from the one or more clients associated with the FL process, the second message indicating that the corresponding client is exiting the FL process.

[0251] Additionally or alternatively, computer program 1710 may be configured as computer program code implemented in computer program modules 1710C and 1710D. Thus, in an exemplary embodiment, when arrangement 1700 is used in a server associated with a FL process, the code in the computer program of arrangement 1700 includes: module 1710C configured to receive one or more fifth messages indicating analysis data associated with one or more clients in the FL process and / or one or more candidate clients to be selected for the FL process; and module 1710D configured to select at least one client for the FL process from the one or more candidate clients and / or the one or more clients based at least on the analysis data.

[0252] Additionally or alternatively, the computer program 1710 may be configured as computer program code implemented in computer program modules 1710E and / or 1710F. Thus, in an exemplary embodiment, when the arrangement 1700 is used in a client associated with a FL process, the code in the computer program of the arrangement 1700 includes at least one of the following: a module 1710E configured to receive a first message from a server associated with the FL process, the first message indicating that the client has been deselected for the FL process by the server and / or that the FL process is suspended; and a module 1710F configured to send a second message to the server associated with the FL process, the second message indicating that the client is exiting the FL process.

[0253] Additionally or alternatively, the computer program 1710 may be configured as computer program code implemented in computer program modules 1710G and / or 1710H. Thus, in an exemplary embodiment, when the arrangement 1700 is used in a client associated with a FL process or a candidate client to be selected for the FL process, the code in the computer program of the arrangement 1700 includes at least one of the following: a module 1710G configured to send a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with the client or candidate client; and a module 1710H configured to send a seventh message to one or more network nodes, the seventh message indicating data associated with the client or candidate client, the data being used as input data when determining analysis data associated with the client or candidate client.

[0254] Additionally or alternatively, the computer program 1710 may be configured as computer program code structured in a computer program module 1710I. Thus, in an exemplary embodiment, when the arrangement 1700 is used in a network node, the code in the computer program of the arrangement 1700 includes a module 1710I configured to send a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with one or more clients associated with the FL process and / or one or more candidate clients to be selected by the server for the FL process.

[0255] A computer program module is essentially executable Figures 2 to 16 The actions of the process shown in the embodiment of the present invention are used to simulate a server, a client and / or a network node. In other words, when different computer program modules are executed in the processing unit 1706, these computer program modules may correspond to different modules in the server, the client and / or the network node.

[0256] Although the above combination Figure 17 The code device in the disclosed embodiment is implemented as a computer program module, which, when executed in a processor, causes the device to perform the actions described above in conjunction with the above figures. In an alternative embodiment, at least one code device can be at least partially implemented as a hardware circuit.

[0257] The processor may be a single CPU (central processing unit), but may also include two or more processing units. For example, the processor may include a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC). The processor may also include onboard memory for caching purposes. The computer program may be carried by a computer program product connected to the processor. The computer program product may include a computer-readable medium on which the computer program is stored. For example, the computer program product may be a flash memory, a random access memory (RAM), a read-only memory (ROM), or an EEPROM, and the above-mentioned computer program modules may, in alternative embodiments, be distributed on different computer program products in the form of memory within a server, a client, and / or a network node.

[0258] An exemplary server associated with the FL process is provided corresponding to the method 1200 described above. Figure 18 FIG1 is a block diagram of a server 1800 according to an embodiment of the present disclosure. In some embodiments, the server 1800 may be, for example, the server NWDAF 120-S.

[0259] The server 1800 may be configured to perform the above combined Figure 12 The method 1200 is described. Figure 18As shown, the server 1800 may include at least one of the following items: a sending module 1810, configured to send a first message to one or more clients associated with the FL process, the first message indicating that the corresponding client is deselected by the server for the FL process and / or the FL process is suspended; and a receiving module 1820, configured to receive a second message from one or more clients associated with the FL process, the second message indicating that the corresponding client is exiting the FL process.

[0260] The modules 1810 and / or 1820 may be implemented as a pure hardware solution or as a combination of software and hardware, for example, by one or more of the following: a module configured to perform the above description and, for example, Figure 12 The server 1800 may further include one or more additional modules, each of which may execute the operations described in the preceding text. Figure 12 Any step of method 1200 as described.

[0261] An exemplary server associated with the FL process is provided corresponding to the method 1300 described above. Figure 19 FIG. 1 is a block diagram of a server 1900 according to an embodiment of the present disclosure. In some embodiments, the server 1900 may be, for example, the server NWDAF 120-S.

[0262] The server 1900 may be configured to perform the above combined Figure 13 The method 1300 is described. Figure 19 As shown, the server 1900 may include: a receiving module 1910, configured to receive one or more fifth messages, the one or more fifth messages indicating analysis data associated with one or more clients in the FL process and / or one or more candidate clients to be selected for the FL process; and a selection module 1920, configured to select at least one client for the FL process from the one or more candidate clients and / or the one or more clients based at least on the analysis data.

[0263] The modules 1910 and 1920 can be implemented as a pure hardware solution or as a combination of software and hardware, for example, by one or more of the following: Figure 13 The server 1900 may further include one or more additional modules, each of which may execute the operations described in the referenced embodiment. Figure 13Any step of method 1300 as described.

[0264] An exemplary client associated with the FL process is provided corresponding to the method 1400 described above. Figure 20 2 is a block diagram of a client 2000 according to an embodiment of the present disclosure. In some embodiments, the client 2000 may be, for example, the client NWDAF 120-C.

[0265] Client 2000 can be configured to perform the above combined Figure 14 The method 1400 is described. Figure 20 As shown, the client 2000 may include at least one of the following items: a receiving module 2010, configured to receive a first message from a server associated with the FL process, the first message indicating that the client is deselected by the server for the FL process and / or the FL process is suspended; and a sending module 2020, configured to send a second message to the server associated with the FL process, the second message indicating that the client is exiting the FL process.

[0266] The modules 2010 and / or 2020 may be implemented as a pure hardware solution or as a combination of software and hardware, for example, by one or more of the following: configured to perform the above-described and, for example, Figure 14 The client 2000 may further comprise one or more additional modules, each of which may execute the operations described in the reference. Figure 14 Any step of method 1400 as described.

[0267] Corresponding to the method 1500 described above, exemplary clients associated with the FL process or candidate clients to be selected for the FL process are provided. Figure 21 is a block diagram of a client or candidate client 2100 according to an embodiment of the present disclosure. In some embodiments, the client or candidate client 2100 may be, for example, the client NWDAF 120-C.

[0268] The client or candidate client 2100 may be configured to perform the above combined Figure 15 The method 1500 is described. Figure 21As shown, the client or candidate client 2100 may include at least one of the following items: a first sending module 2110, configured to send a fifth message to a server associated with the FL process, the fifth message indicating analysis data associated with the client or candidate client; and a second sending module 2120, configured to send a seventh message to one or more network nodes, the seventh message indicating data associated with the client or candidate client, which data is used as input data when determining the analysis data associated with the client or candidate client.

[0269] The modules 2110 and / or 2120 may be implemented as a pure hardware solution or as a combination of software and hardware, for example, by one or more of the following: Figure 15 A processor or microprocessor and appropriate software and a memory, PLD or other electronic component or processing circuit for storing the software. In addition, the client or candidate client 2100 may include one or more additional modules, each of which may perform the reference Figure 15 Any step of method 1500 as described.

[0270] An exemplary network node is provided corresponding to the method 1600 described above. Figure 22 2 is a block diagram illustrating a network node 2200 according to an embodiment of the present disclosure. In some embodiments, the network node 2200 may be, for example, a secondary NWDAF 120-A.

[0271] The network node 2200 may be configured to perform the above combined Figure 16 The method 1600 is described. Figure 22 As shown, the network node 2200 may include a sending module 2210, which is configured to send a fifth message to a server associated with the FL process, wherein the fifth message indicates analysis data associated with the following items: one or more clients associated with the FL process and / or one or more candidate clients to be selected by the server for the FL process.

[0272] The module 2210 can be implemented as a pure hardware solution or as a combination of software and hardware, for example, it can be implemented by one or more of the following: Figure 16 A processor or microprocessor and appropriate software and a memory, PLD or other electronic component or processing circuit for storing the actions shown in FIG. In addition, the network node 2200 may include one or more additional modules, each of which may perform the operations described in FIG. Figure 16 Any step of method 1600 as described.

[0273] The present disclosure has been described above with reference to the embodiments of the present disclosure. However, these embodiments are for illustrative purposes only and are not intended to limit the present disclosure. The scope of the present disclosure is defined by the appended claims and their equivalents. A person skilled in the art may make various modifications and variations without departing from the scope of the present disclosure, and such modifications and variations are intended to fall within the scope of the present disclosure.

[0274] Explanation of abbreviations

[0275] 5GC5G core network

[0276] AF application function

[0277] AI Artificial Intelligence

[0278] DML distributed machine learning

[0279] FL Federated Learning

[0280] ML Machine Learning

[0281] MTLF model training logic function

[0282] NEF network exposure function

[0283] NF Network Function

[0284] NRF Network Repository Features

[0285] NWDAF network data analysis function

[0286] OAM operation management and maintenance.

Claims

1. A method (1200) at a server (120-S) associated with a federated learning (FL) process, the method (1200) comprising at least one of the following: sending (S1210) a first message to one or more clients (120-C) associated with the FL process, the first message indicating that the corresponding client is deselected by the server (120-S) for the FL process and / or that the FL process is suspended; and A second message is received (S1220) from one or more clients (120-C) associated with the FL process, the second message indicating that the corresponding client is exiting the FL process.

2. The method (1200) according to claim 1, wherein At least one of the first message and the second message further indicates at least one of the following: FL related identifier ID; and Reason code.

3. The method (1200) of claim 2, wherein: The reason code indicates at least one of the following: When the reason code is indicated by the first message, the corresponding client is deselected by the server (120-S) for the FL process; When the reason code is indicated by the first message, the FL process is suspended; When the reason code is indicated by the corresponding second message, an availability change associated with the corresponding client; and When the reason code is indicated by the corresponding second message, capabilities associated with the corresponding client change.

4. The method (1200) according to any one of claims 1 to 3, further comprising at least one of the following: receiving, in response to the corresponding first message, a third message from at least one of the one or more clients (120-C), the third message indicating that the FL process is terminated or is to be terminated at the at least one client; and In response to the corresponding second message, a fourth message is sent to at least one of the one or more clients (120-C), the fourth message indicating that the at least one client is removed or is to be removed from the FL process.

5. The method (1200) according to claim 4, wherein The fourth message also indicates the time when the at least one client exits the FL process.

6. The method (1200) according to any one of claims 1 to 5, further comprising at least one of the following: In response to sending the corresponding first message, stopping one or more FL operations of the FL process associated with at least one of the one or more clients (120-C); in response to receiving the corresponding second message, stopping one or more FL operations of the FL process associated with at least one of the one or more clients (120-C); and In response to receiving the corresponding third message, one or more FL operations of the FL process associated with at least one of the one or more clients (120-C) are stopped.

7. The method (1200) according to any one of claims 1 to 6, wherein: The first message causes the client (120-C) to stop one or more FL operations of the FL process.

8. The method (1200) according to any one of claims 1 to 7, wherein: The first message is one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Terminate request message, and / or The third message is a corresponding message among the Nnwdaf_MLModelProvision_Unsubscribe request message, the Nnwdaf_MLModelProvision_Unsubscribe response message, and the Nnwdaf_MLTraining_Terminate response message.

9. The method (1200) according to any one of claims 1 to 8, wherein The second message is one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Quit request message, and / or The fourth message is a corresponding message among the Nnwdaf_MLModelProvision_Unsubscribe request message, the Nnwdaf_MLModelProvision_Unsubscribe response message, and the Nnwdaf_MLTraining_Quit response message.

10. The method (1200) according to any one of claims 1 to 9, wherein Before the step of sending (S1210) the first message and / or receiving (S1220) the second message, the method (1200) further includes at least one of the following: Sending machine learning ML model information to the one or more clients (120-C); and ML model information is received from the one or more clients (120-C).

11. The method (1200) according to any one of claims 1 to 10, wherein: The network data analysis function NWDAF is hosted by the server (120-S), and / or One or more NWDAFs are hosted by the one or more clients (120-C).

12. The method (1200) according to any one of claims 1 to 11, further comprising any step of the method (1300) according to any one of claims 13 to 21.

13. A method (1300) at a server (120-S) associated with a FL process, the method (1300) comprising: receiving (S1310) one or more fifth messages indicating analysis data associated with one or more clients (120-C) in the FL process and / or one or more candidate clients (120-C) to be selected for the FL process; and Based on at least the analysis data, at least one client is selected ( S1320 ) from the one or more candidate clients and / or the one or more clients for the FL process.

14. The method (1300) according to claim 13, wherein: The one or more fifth messages are received from at least one of: at least one client among the one or more clients (120-C); at least one candidate client of the one or more candidate clients (120-C); and One or more network nodes (120-A).

15. The method (1300) of claim 14, wherein: The one or more network nodes (120-A) are network nodes that assist the server (120-S) in collecting analytical data associated with at least one of the one or more clients (120-C) and / or at least one of the one or more candidate clients (120-C).

16. The method (1300) according to any one of claims 13 to 15, wherein Before the step of receiving (S1310) the one or more fifth messages, the method (1300) further comprises at least one of the following: Sending a sixth message to at least one of the one or more clients (120-C) and / or at least one candidate client of the one or more candidate clients, the sixth message being for subscribing to analysis data associated with the at least one client and / or the at least one candidate client; as well as A sixth message is sent to at least one of the one or more network nodes (120-A), the sixth message being for subscribing to analysis data associated with at least one of the one or more clients (120-C) and / or at least one of the one or more candidate clients (120-C).

17. The method (1300) according to any one of claims 13 to 16, wherein The fifth message is a Nnwdaf_AnalyticsSubscription_Notify request message, The sixth message is a Nnwdaf_AnalyticsSubscription_Subscribe request message.

18. The method (1300) according to any one of claims 13 to 17, wherein The analytical data includes data related to at least one of the following: load; availability; ability; Latency; and accuracy.

19. The method (1300) according to any one of claims 13 to 18, wherein The step of receiving ( S1310 ) the one or more fifth messages is performed periodically and / or dynamically in response to an event.

20. The method (1300) according to any one of claims 13 to 19, wherein The network data analysis function NWDAF is hosted by the server (120-S), and / or wherein the NWDAF is hosted by the one or more clients (120-C), and / or The NWDAF is hosted by the one or more candidate clients (120-C).

21. The method (1300) according to any one of claims 13 to 20, further comprising any step of the method (1200) according to any one of claims 1 to 12.

22. A server (120-S, 1700, 1800, 1900), comprising: Processor(1706); A memory (1708) storing instructions which, when executed by the processor (1706), cause the processor (1706) to perform the method (1200, 1300) according to any one of claims 1 to 21.

23. A method (1400) at a client (120-C) associated with an FL process, the method (1400) comprising at least one of the following: receiving (S1410) a first message from a server (120-S) associated with the FL process, the first message indicating that the client (120-C) is deselected by the server (120-S) for the FL process and / or that the FL process is suspended; and A second message is sent (S1420) to a server (120-S) associated with the FL process, the second message indicating that the client (120-C) is exiting the FL process.

24. The method (1400) of claim 23, wherein: At least one of the first message and the second message further indicates at least one of the following: FL related identifier ID; and Reason code.

25. The method (1400) of claim 24, wherein the reason code indicates at least one of: When the reason code is indicated by the first message, the client (120-C) is deselected by the server (120-S) for the FL process; When the reason code is indicated by the first message, the FL process is suspended; an availability change associated with the client (120-C) when the reason code is indicated by the second message; and When the reason code is indicated by the second message, capabilities associated with the client (120-C) change.

26. The method (1400) according to any one of claims 23 to 25, further comprising at least one of the following: In response to the first message, sending a third message to the server (120-S), the third message indicating that the FL process is terminated or is about to be terminated at the client (120-C); and In response to the second message, a fourth message is received from the server (120-S), the fourth message indicating that the client (120-C) is removed or is to be removed from the FL process.

27. The method (1400) of claim 26, wherein: The fourth message also indicates the time when the client (120-C) exits the FL process.

28. The method (1400) according to any one of claims 23 to 27, further comprising at least one of the following: In response to receiving the first message, stopping one or more FL operations of the FL process; In response to receiving the first message, clearing the context of the FL process; In response to sending the second message, stopping one or more FL operations of the FL process; In response to sending the second message, clearing the context of the FL process; as well as In response to receiving the fourth message, stopping one or more FL operations of the FL process; In response to receiving the fourth message, the context of the FL process is cleared.

29. The method (1400) according to any one of claims 23 to 28, wherein The first message is one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Terminate request message, and / or The third message is a corresponding message among the Nnwdaf_MLModelProvision_Unsubscribe request message, the Nnwdaf_MLModelProvision_Unsubscribe response message, and the Nnwdaf_MLTraining_Terminate response message.

30. The method (1400) according to any one of claims 23 to 29, wherein the second message is one of a Nnwdaf_MLModelProvision_Unsubscribe request message and a Nnwdaf_MLTraining_Quit request message, and / or in, The fourth message is a corresponding message among the Nnwdaf_MLModelProvision_Unsubscribe request message, the Nnwdaf_MLModelProvision_Unsubscribe response message, and the Nnwdaf_MLTraining_Quit response message.

31. The method (1400) according to any one of claims 23 to 30, wherein Before the step of receiving (S1410) the first message and / or sending (S1420) the second message, the method (1400) further includes at least one of the following: receiving machine learning (ML) model information from the server (120-S); and The ML model information is sent to the server (120-S).

32. The method (1400) according to any one of claims 23 to 31, wherein The network data analysis function NWDAF is hosted by the client (120-C), and / or The NWDAF is hosted by the server (120-S).

33. The method (1400) according to any one of claims 23 to 32, further comprising any step of the method (1500) according to any one of claims 34 to 42.

34. A method (1500) at a client (120-C) associated with a FL process or at a candidate client (120-C) to be selected for the FL process, the method (1500) comprising at least one of: sending (S1510) a fifth message to a server (120-S) associated with the FL process, the fifth message indicating analysis data associated with the client (120-C) or the candidate client (120-C); A seventh message is sent (S1520) to one or more network nodes (120-A), the seventh message indicating data associated with the client (120-C) or the candidate client (120-C), the data being used as input data when determining analysis data associated with the client (120-C) or the candidate client (120-C).

35. The method (1500) of claim 34, wherein: The one or more network nodes (120-A) are network nodes that assist the server (120-S) in collecting analytical data associated with the client (120-C) or the candidate client (120-C).

36. The method (1500) of claim 34 or 35, wherein: Before the step of sending (S1510) the fifth message, the method (1500) further includes at least one of the following items: receiving a sixth message from the server (120-S), the sixth message being for subscribing to analysis data associated with the client (120-C) or the candidate client (120-C); and An eighth message is received from at least one of the one or more network nodes (120-A), the eighth message requesting input data associated with the client (120-C) or the candidate client (120-C).

37. The method (1500) according to any one of claims 34 to 36, further comprising: collecting data for determining analytical data associated with the client (120-C) or the candidate client (120-C) from at least one of one or more network functions NF, one or more application functions AF, and one or more operations administration and maintenance OAM nodes; and Analysis is performed on the collected data to determine analytical data associated with the client (120-C) or the candidate client (120-C).

38. The method (1500) according to any one of claims 34 to 37, wherein The fifth message is a Nnwdaf_AnalyticsSubscription_Notify request message, The sixth message is a Nnwdaf_AnalyticsSubscription_Subscribe request message.

39. The method (1500) according to any one of claims 34 to 38, wherein The analytical data includes data related to at least one of the following: load; availability; ability; Latency; and accuracy.

40. The method (1500) according to any one of claims 34 to 39, wherein The step of sending (S1510) the fifth message is performed periodically and / or dynamically in response to an event.

41. The method (1500) according to any one of claims 34 to 40, wherein A network data analysis function NWDAF is hosted by the client (120-C) or the candidate client (120-C), and / or The NWDAF is hosted by the server (120-S).

42. The method (1500) according to any one of claims 34 to 41, further comprising any step of the method (1400) according to any one of claims 23 to 33.

43. A client (120-C, 1700, 2000, 2100), comprising: Processor(1706); A memory (1708) storing instructions which, when executed by the processor (1706), cause the processor (1706) to perform the method (1400, 1500) according to any one of claims 23 to 42.

44. A method (1600) at a network node (120-A), the method (1600) comprising: A fifth message is sent (S1610) to a server (120-S) associated with a FL process, the fifth message indicating analysis data associated with one or more clients (120-C) associated with the FL process and / or one or more candidate clients (120-C) to be selected by the server (120-S) for the FL process.

45. The method (1600) of claim 44, wherein: The network node (120-A) is a network node that assists the server (120-S) in collecting analytical data associated with the one or more clients (120-C) and / or the one or more candidate clients (120-C).

46. The method (1600) of claim 44 or 45, wherein Before the step of sending (S1610) the fifth message, the method (1600) further includes: A sixth message is received from the server (120-S), the sixth message being for subscribing to analytical data associated with the one or more clients (120-C) and / or the one or more candidate clients (120-C).

47. The method (1600) according to any one of claims 44 to 46, wherein The fifth message is a Nnwdaf_AnalyticsSubscription_Notify request message, The sixth message is a Nnwdaf_AnalyticsSubscription_Subscribe request message.

48. The method (1600) according to any one of claims 44 to 47, wherein Before the step of sending (S1610) the fifth message, the method (1600) further includes: collecting data for determining analytical data associated with the one or more clients (120-C) and / or the one or more candidate clients (120-C) from at least one of the one or more clients (120-C), the one or more candidate clients (120-C), one or more network functions NF, one or more application functions AF, and one or more operations administration and maintenance OAM nodes; and Analysis is performed on the collected data to determine analytical data associated with the one or more clients (120-C) and / or the one or more candidate clients (120-C).

49. The method (1600) according to any one of claims 44 to 48, wherein The analytical data includes data related to at least one of the following: load; availability; ability; Latency; and accuracy.

50. The method (1600) according to any one of claims 44 to 49, wherein The step of sending (S1610) the fifth message is performed periodically and / or dynamically in response to an event.

51. The method (1600) according to any one of claims 44 to 50, wherein NWDAF is hosted by the client (120-C) and / or the candidate client (120-C), and / or The NWDAF is hosted by the server (120-S).

52. A network node (120-A, 1700, 2200), comprising: Processor(1706); A memory (1708) storing instructions which, when executed by the processor (1706), cause the processor (1706) to perform the method (1600) according to any one of claims 44 to 51.

53. A computer program (1710) comprising instructions which, when executed by at least one processor (1706), cause the at least one processor (1706) to perform the method according to any one of claims 1 to 21, 23 to 42 and 44 to 51.

54. A carrier (1708) comprising a computer program (1710) according to claim 53, wherein The carrier (1708) is one of an electric signal, an optical signal, a radio signal or a computer-readable storage medium.

55. A telecommunication system (10) for supporting federated learning (FL), the telecommunication system (10) comprising: The server (120-S) according to claim 22; and One or more clients (120-C) according to claim 43.

56. The telecommunication system (10) according to claim 55, further comprising: One or more network nodes (120-A) according to claim 52.

Images