Management method of machine learning model for network data analytics function device
View Patent ↗A machine learning (ML) model management method for a network data analytics function (NWDAF) device is disclosed. The NWDAF device performs at least one of an analytics logical function (AnLF) for network data and an ML model training logical function (MTLF).
1. A method for discovering a machine learning (ML) model, the method performed by a first network data analytics function (NWDAF) device and comprising:
invoking, from a network repository function (NRF) device, an ML model discovery request service operation;
receiving, from the NRF device in response to the ML model discovery request service operation, a discovery response including information for each of one or more NWDAF instances, the information for each NWDAF instance including ML model-related information provided by a corresponding NWDAF device that performs a model training logical function (MTLF);
selecting an NWDAF instance from the one or more NWDAF instances included in the discovery response,
provisioning, to the first NWDAF using a second NWDAF device corresponding to the selected NWDAF instance, an ML model;
analyzing network data using the ML model.
2. The method of claim 1 , wherein
the NRF device is configured to store a network function (NF) profile for the MTLF by invoking, from the second NWDAF device, a registration request service operation with an NF, and
the registration request service operation includes at least one of (i) a list of Analytic IDs, (ii) a supported service, (iii) a serving area, and (iv) subscribed network slice selection assistance information (S-NSSAI).
3. The method of claim 1 , wherein
the NRF device is configured to store an NF profile for the MTLF by invoking, from the second NWDAF device, a registration request service operation with an NF, and
the registration request service operation includes at least one of (i) a list of Analytic IDs, (ii) a supported service, (iii) a serving area, (iv)) subscribed network slice selection assistance information (S-NSSAI), and (v) ML model information including at least one of an ML model file address, an ML model file, a model ID, and a model version.
4. The method of claim 1 , wherein the selecting of the NWDAF instance comprises selecting an NWDAF instance based on at least one of (i) subscribed network slice selection assistance information (S-NSSAI), (ii) an Analytic ID, (iii) a supported service, (iv) NWDAF serving area information, (v) NWDAF location information, (vi) an NF type of a data source, (vii) an NF Set ID of the data source, (viii) a supported analytics delay, and (ix) an NWDAF capability.
5. The method of claim 1 , wherein
the first NWDAF device is configured to perform local training of federation learning, and support the MTLF, and
the second NWDAF device is configured to perform global training of federation learning, and support the MTLF.
6. The method of claim 5 , wherein the registration request service operation includes at least one of a list of Analytic IDs, a supported service, a serving area, subscribed network slice selection assistance information (S-NSSAI), ML model information including at least one of an ML model file address, an ML model file, a model ID, and a model version, and an ML model training capability or an ML model update capability.
7. A method for provisioning an a machine learning (ML) model, the method performed by a first network data analytics function (NWDAF) device and comprising:
invoking, from a second NWDAF device that performs a model training logical function (MTLF), a subscription service operation for provisioning of the ML model;
invoking, from the second NWDAF device, a notification service operation for the subscription service;
provisioning, to the first NWDAF using the second NWDAF device, the ML model;
analyzing network data using the ML model.
8. The method of claim 7 , wherein the subscription service operation includes at least one of (i) an Analytic ID, (ii) subscribed network slice selection assistance information (S-NSSAI), (iii) a target area of interest, (iv) an application ID, (v) a target user equipment (UE), (v) an ML model target period, and (vi) an expiry time.
9. The method of claim 7 , wherein the subscription service operation includes at least one of (i) an Analytic ID, (ii) subscribed network slice selection assistance information (S-NSSAI), (iii) a target area of interest, (iv) an application ID, (v) a target UE, (v) an ML model target period, (vi) an expiry time, and (vii) ML model information including at least one of an ML model file address, an ML model file, a model ID, and a model version.
10. The method of claim 7 , wherein the notification service operation includes at least one of (i) ML model information including an ML model file or an ML model file address, (ii) a validity period, and (iii) a spatial validity.
11. The method of claim 7 , wherein the notification service operation includes at least one of (i) ML model information including at least one of an ML model file address, an ML model file, a model ID, and a model version, (ii) a validity period, and (iii) a spatial validity.
12. The method of claim 7 , wherein
the invoking of the subscription service operation for provisioning of the ML model comprises invoking a subscription service operation for provisioning of a second ML model after a subscription for provisioning of a first ML model is completed, and
the subscription service operation includes a subscription ID same as a subscription ID for the first ML model.
13. The method of claim 12 , wherein the subscription service operation includes at least one of (i) an Analytic ID, (ii) subscribed network slice selection assistance information (S-NSSAI), (iii) a target area of interest, (iv) an application ID, (v) a target UE, (v) an ML model target period, (vi) an expiry time, (vii) ML model information including at least one of an ML model file address, an ML model file, a model ID, and a model version, and (viii) an alternative ML model flag.
14. The method of claim 7 , wherein
the first NWDAF device is configured to perform local training of federation learning, and support the MTLF, and
the second NWDAF device is configured to perform global training of federation learning, and support the MTLF.
15. The method of claim 14 , wherein the notification service operation includes at least one of (i) ML model information including at least one of an ML model file address, an ML model file, a model ID, and a model version, (ii) a validity period, and (iii) a spatial validity.
16. The method of claim 7 , further comprising:
locally training the ML model;
invoking, from the second NWDAF device, an ML model update notification service operation; and
invoking, from the second NWDAF device that globally updates the ML model, a notification service operation for provisioning of the ML model,
wherein the first NWDAF device is configured to perform local training of federation learning on the ML model, and
the second NWDAF device is configured to perform global training of federation learning on the ML model.
17. The method of claim 7 , further comprising:
locally training the ML model; and
invoking, from the second NWDAF device that globally updates the ML model, an ML model update notification service operation,
wherein the first NWDAF device is configured to perform local training of federation learning on the ML model, and
the second NWDAF device is configured to perform global training of federation learning on the ML model.