Network data analysis method and system based on federated learning

Patent metadata

Field	Value
Publication number	US-12309038-B2
Application number	US-202318313205-A
Country	US
Kind code	B2
Filing date	May 5, 2023
Priority date	May 6, 2022
Publication date	May 20, 2025
Grant date	May 20, 2025

Abstract

A federated learning-based network data analysis method and an apparatus for performing the same are provided. An operating method of a federation learning (FL) server includes performing an FL operation trigger in response to a request of a first network function (NF), selecting a plurality of second NFs on an analytics identifier (ID) in response to the FL operation trigger, and requesting the plurality of second FLs for FL, wherein the first NF is an FL consumer, and the plurality of second NFs is an FL client.

First claim

What is claimed is:

1. An operating method of a federation learning (FL) server, the method comprising:
performing an FL operation trigger in response to a request of a first network function (NF); and
in response to the FL operation trigger:
selecting a plurality of second NFs on an analytics identifier (ID);
requesting the plurality of second NFs for FL,
receiving one or more trained interim machine learning (ML) models from the plurality of second NFs,
aggregating the trained interim ML models from the plurality of second NFs to produce a trained ML model, and
providing the trained ML model to the first NF,

wherein the first NF is an FL consumer, and
wherein the plurality of second NFs includes an FL client and does not include the first NF.

2. The method of claim 1 , wherein the FL server and the plurality of second NFs comprise an FL capability, and
the FL capability indicates that at least one of an FL server capability and an FL client capability is supported for a corresponding analytics ID.

3. The method of claim 1 , wherein, when the first NF is a network data analytics function (NWDAF) comprising a model training logical function (MTLF), a request of the first NF is performed as an Nnwdaf_MLModelTraining_Subscribe service, and
when the first NF is an NWDAF comprising an analytic logical function (AnLF), a request of the first NF is performed as an Nnwdaf_MLModelProvsion_Subscribe service.

4. The method of claim 1 , wherein the requesting for FL comprises:
transmitting, to the plurality of second NFs, second FL parameters, which are different from first FL parameters, by the FL server comprising the first FL parameters.

5. The method of claim 4 , wherein the first FL parameters comprise a number of FL rounds, a total number of FL clients used in a process, and an area of interest for an analytics ID.

6. The method of claim 4 , wherein the second FL parameters comprise a machine learning (ML) identifier, an indication for enabling local training, or ML model information for requesting an ML model operation.

7. The method of claim 6 , wherein different ML IDs are assigned to the plurality of second NFs, respectively.

8. The method of claim 1 , wherein the interim ML models are identified by an analytics ID and machine learning (ML) IDs respectively assigned to the plurality of second NFs.

9. The method of claim 1 , further comprising:
determining whether an FL operation is required based on an area of interest, machine learning (ML) model reporting information, an expiry time, and a local operator configuration.

10. A federation learning (FL) server apparatus comprising:
a processor; and
a memory electrically connected to the processor and configured to store instructions executable by the processor,
wherein the processor performs a plurality of operations when the instructions are executed by the processor, the plurality of operations comprising:
performing an FL operation trigger in response to a request of a first network function (NF);
in response to the FL operation trigger:
selecting a plurality of second NFs on an analytics identifier (ID) in response to the FL operation trigger;
requesting the plurality of second NFs for FL;
receiving one or more trained interim machine learning (ML) models from the plurality of second NFs,
aggregating the trained interim ML models from the plurality of second NFs to produce a trained ML model, and
providing the trained ML model to the first NE,

wherein the first NF is an FL consumer, and
wherein the plurality of second NFs includes an FL client and does not include the first NF.

11. The FL server apparatus of claim 10 , wherein the FL server apparatus and the plurality of second NFs comprise an FL capability, and
the FL capability indicates that at least one of an FL server capability and an FL client capability is supported for a corresponding analytics ID.

12. The FL server apparatus of claim 10 , wherein, when the first NF is a network data analytics function (NWDAF) comprising a model training logical function (MTLF), a request of the first NF is performed as an Nnwdaf_MLModelTraining_Subscribe service, and
when the first NF is an NWDAF comprising an analytic logical function (AnLF), a request of the first NF is performed as an Nnwdaf_MLModelProvsion_Subscribe service.

13. The FL server apparatus of claim 10 , wherein the requesting for FL comprises:
transmitting, to the plurality of second NFs, second FL parameters, which are different from FL parameters, by the FL server apparatus comprising the first FL parameters.

14. The FL server apparatus of claim 13 , wherein the first FL parameters comprise a number of FL rounds, a total number of FL clients used in a process, and an area of interest for an analytics ID.

15. The FL server apparatus of claim 13 , wherein the second FL parameters comprise a machine learning (ML) identifier, an indication for enabling local training, or ML model information for requesting an ML model operation.

16. The FL server apparatus of claim 15 , wherein different ML IDs are assigned to the plurality of second NFs, respectively.

17. The FL server apparatus of claim 11 , wherein the interim ML models are identified by an analytics ID and the ML IDs respectively assigned to the plurality of second NFs.

18. The FL server apparatus of claim 10 , wherein the plurality of operations further comprises:
determining whether an FL operation is required based on an area of interest, machine learning (ML) model reporting information, an expiry time, and a local operator configuration.

Assignees

• Electronics & Telecommunications Res Inst

Inventors

• Lee Soohwan

Classifications

• H04L41/16

  using machine learning or artificial intelligence · CPC title

• G06N20/20

  Ensemble learning · CPC title

• H04W24/02

  Arrangements for optimising operational condition · CPC title

• G06N20/00

  Machine learning · CPC title

• H04L41/40

  using virtualisation of network functions or resources, e.g. SDN or NFV entities · CPC title