Apparatus, method and computer program

Patent metadata

Field	Value
Publication number	US-12407580-B2
Application number	US-202418432149-A
Country	US
Kind code	B2
Filing date	Feb 5, 2024
Priority date	Feb 10, 2023
Publication date	Sep 2, 2025
Grant date	Sep 2, 2025

Abstract

There is provided an apparatus means for, at a central node associated with a plurality of distributed nodes, determining that at least one distributed node of the plurality of distributed nodes has not provided training information relating to a training process of a machine learning model before expiry of a first timer for a given iteration of N iterations of the training process, means for generating analytic information relative to the at least one distributed node, wherein the analytic information comprises a count based on the determining and means for providing the analytic information to a storage function.

First claim

The invention claimed is:

1. An apparatus for a communication network, the apparatus comprising:
at least one processor; and
at least one memory storing instructions, which when executed by the at least one processor, cause the apparatus at least to perform:
determining that at least one distributed node of a plurality of distributed nodes associated with the apparatus has not provided training information relating to a training process of a machine learning model before expiry of a first timer for a given iteration of N iterations of the training process;
generating analytic information relative to the at least one distributed node, wherein the analytic information comprises a count based on the determining; and
providing the analytic information to a storage function.

2. The apparatus according to claim 1 , wherein the instructions, when executed by the at least one processor, further cause the apparatus at least to perform:
requesting the at least one distributed node to indicate an expected time to provide the training information.

3. The apparatus according to claim 2 , wherein the instructions, when executed by the at least one processor, further cause the apparatus at least to perform:
receiving, from the at least one distributed node, in response to the requesting, the expected time to provide the training information.

4. The apparatus according to claim 3 , wherein the instructions, when executed by the at least one processor, further cause the apparatus at least to perform:
determining whether the expected time is less than a time associated with a second timer; and
if the expected time is less than the time associated with the second timer, receiving the training information before expiry of the second timer, or
if the expected time is not less than the time associated with the second timer, providing an indication to the at least one distributed node to not provide the training information for the given iteration.

5. The apparatus according to claim 1 , wherein the instructions, when executed by the at least one processor, further cause the apparatus at least to perform:
determining a confidence value for the at least one distributed node based on the analytic information; and
providing the confidence value to the storage function.

6. The apparatus according to claim 1 , wherein the apparatus comprises a root federated learning aggregator, and wherein the plurality of distributed nodes comprises a plurality of federated learning distributed nodes.

7. The apparatus according to claim 1 , wherein the apparatus comprises a root federated learning aggregator, and wherein the plurality of distributed nodes comprises a plurality of lower-level federated learning aggregators.

8. The apparatus according to claim 1 , wherein the analytic information further comprises at least one of the following:
average training information response time for the at least one distributed node over the N iterations,
average delay for providing the training information for the at least distributed node over the N iterations,
a relative degree of delay for providing the training information for the at least distributed node over the N iterations,
a number of iterations performed by the at least one distributed node during the training process,
a number of times the at least one distributed node left and joined the plurality of distributed nodes during the training process,
a number of distributed nodes of the plurality of distributed nodes which participated in at least one of the N iterations of the training process,
a number of distributed nodes of the plurality of distributed nodes which have participated in each iteration of the N iterations of the training process, or
a number of distributed nodes which have joined and left the plurality of distributed nodes during the training process.

9. The apparatus according to claim 1 , wherein the storage function comprises:
an analytics data repository function,
a network repository function, or
an operations, administration and management entity.

10. The apparatus according to claim 1 ,
wherein the apparatus comprises at least one of the following:
a radio access network (RAN) operations, administration and management entity,
a core network (CN) operations, administration and management entity,
a network function for performing data analytics,
a user equipment,
a base station,
a non-real time RAN intelligence controller (RIC), or
a near real time RIC,

and wherein the plurality of distributed nodes comprises at least one of the following:
another RAN operations, administration and management entity,
another CN operations, administration and management entity,
another network function for performing data analytics,
another user equipment,
another base station,
another non-real time RIC, or
another near real time RIC.

11. An apparatus comprising:
at least one processor; and
at least one memory storing instructions, which when executed by the at least one processor, cause the apparatus at least to perform:
receiving, from a central node, a training configuration and a first timer for a given iteration of N iterations of a training process of a machine learning model;
generating training information relating to the training process; and
providing the training information to the central node for use in determining analytic information relative to the apparatus.

12. The apparatus according to claim 11 , wherein the instructions, when executed by the at least one processor, further cause the apparatus at least to perform:
receiving, from the central node, when the training information is not provided to the central node before expiration of the first timer, a request to indicate an expected time to provide the training information; and
providing, in response to the request, to the central node, the expected time to provide the training information.

13. The apparatus according to claim 12 , wherein the instructions, when executed by the at least one processor, further cause the apparatus at least to perform:
receiving a second timer or an indication to not provide the information for the given iteration.

14. The apparatus according to claim 11 , wherein the apparatus comprises at least one of the following:
a radio access network (RAN) operations, administration and management entity,
a core network (CN) operations, administration and management entity,
a network function for performing data analytics,
a user equipment,
a base station,
a non-real time RAN intelligence controller (RIC), or
a near real time RIC.

15. The apparatus according to claim 11 , wherein the central node comprises at least one of the following:
another RAN operations, administration and management entity,
another CN operations, administration and management entity,
another network function for performing data analytics,
another user equipment,
another base station,
another non-real time RIC, or
another near real time RIC.

16. A computer readable medium comprising instructions which, when executed by an apparatus for a network node, cause the apparatus to perform at least the following:
determining that at least one distributed node of a plurality of distributed nodes associated with the apparatus has not provided training information relating to a training process of a machine learning model before expiry of a first timer for a given iteration of N iterations of the training process;
generating analytic information relative to the at least one distributed node, wherein the analytic information co

Assignees

• Nokia Technologies Oy

Inventors

• Subramanya Tejas
• Khatibi Sina
• Giust Fabio

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• H04L41/16 Primary

  using machine learning or artificial intelligence · CPC title

• H04L41/14

  Network analysis or design · CPC title

• H04L41/0895

  Configuration of virtualised networks or elements, e.g. virtualised network function or OpenFlow elements · CPC title

• H04L41/0894

  Policy-based network configuration management · CPC title