Granted Patent B2

US 12,506,521 · App. 18/399,306 · Granted Dec 23, 2025

Inventors: Muhammad Karam Shehzad (Massy, FR); Fanny Jardel (Massy, FR); Luca Rose (Massy, FR); Muhammad Majid Butt (Naperville, IL)

Assignee: Nokia Technologies Oy

H04B7/0456H04B17/309

Abstract

Claims (55)

1 . An apparatus comprising:

at least one processor; and

at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to perform:

receiving, from a network node, a request for training data for data augmentation algorithm;

transmitting, to the network node, a first part of a plurality of parts of the training data, wherein properties of data in the first part of the training data reflect distribution of the training data;

wherein the apparatus is caused to perform either:

i) computing a first set of precoding matrix indices based on the training data;

receiving, from the network node, a second set of precoding matrix indices computed based on data samples generated by the network node based on the first part of the training data transmitted to the network node;

comparing the first set of precoding matrix indices and the second set of precoding matrix to verify the data samples generated by the network node; and

based on determining that a number of indices of the second set of precoding matrix indices verified as correct remains below a pre-set threshold, transmitting a second part of the plurality of parts of the training data to the network node; or

ii) before transmitting the first part of the training data, the apparatus is caused to perform:

generating data samples based on one or more parts of the plurality of parts of the training data, wherein properties of data in the one or more parts reflect distribution of the training data;

verifying the generated data samples;

iteratively repeating generating the data samples and verifying the generated data samples to determine a suitable number of the one or more parts of the training data to enable generation of a desired set of training data; and

wherein the first part of the training data to be transmitted to the network node comprises the suitable number of the one or more parts of the training data.

2 . The apparatus of claim 1 , wherein the training data is generated by the apparatus using a known channel estimation algorithm.

3 . The apparatus of claim 1 , wherein the apparatus is further caused to perform:

transmitting, to the network node, an acknowledgement that the apparatus has training data available.

4 . The apparatus of claim 1 , wherein the apparatus is further caused to perform:

dividing the training data into the plurality of parts such that properties of data in the plurality of parts reflect distribution of the training data.

5 . The apparatus of option i of claim 1 , wherein the apparatus is further caused to perform:

based on determining that a number of indices of the second set of precoding matrix indices verified as correct reaches a pre-set threshold, transmitting an acknowledgement to the network node that the data samples generated by the network node are correct.

6 . The apparatus of option i of claim 1 , wherein the apparatus is further caused to perform:

receiving, from the network node, a third set of precoding matrix indices computed based on data samples generated by the network node based on the first part and the second part of the training data transmitted to the network node;

comparing the third set of precoding matrix indices and corresponding precoding matrix indices computed by the apparatus to verify the data samples generated by the network node; and

based on determining that a number of indices of the third set of precoding matrix indices verified as correct remains below a pre-set threshold, transmitting a further part of the one or more parts of the training data to the network node.

7 . The apparatus of option i of claim 1 , wherein the first part comprises more data samples than the second part.

8 . The apparatus of option ii of claim 1 , wherein the generating the data samples is performed using a generative adversarial network, GAN; and wherein the apparatus and the network node have agreed on using twin GAN.

9 . An apparatus comprising:

at least one processor; and

at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to perform:

transmitting, to a user equipment, a request for training data for data augmentation algorithm;

receiving, from the user equipment, a first part of a plurality of parts of the training data, wherein properties of data in the first part of the training data reflect distribution of the training data;

wherein the apparatus is caused to perform either:

i) generating data samples based on the received part of the training data;

computing a set of precoding matrix indices based on the generated data samples;

transmitting the set of precoding matrix indices to the user equipment;

receiving a second part of the plurality of parts of the training data from the user equipment; or

ii) wherein the received part of the training data comprises a suitable number of parts of the training data generated and verified by the user equipment, wherein the suitable number of parts of the training data enables the apparatus to generate a desired set of data using the data augmentation algorithm.

10 . The apparatus of claim 9 , wherein the apparatus is further caused to perform:

receiving, from the user equipment, an acknowledgement that the user equipment has training data available.

11 . The apparatus of claim 9 , wherein the generating the data samples is performed using a generative adversarial network, GAN.

12 . The apparatus of claim option i of claim 9 , wherein the apparatus is further caused to perform:

transmitting a pre-set threshold defining a number of indices of the set of precoding matrix indices that is to be verified as correct.

13 . The apparatus of option i of claim 9 , wherein the apparatus is further caused to perform:

receiving an acknowledgement from the user equipment that the data samples generated by the network node are correct; and

training a machine learning model for a target goal with the data samples.

14 . The apparatus of option i of claim 9 , wherein the apparatus is further caused to perform:

generating further data samples based on the second part of the training data;

computing a further set of precoding matrix indices based on the further data samples;

transmitting the further set of precoding matrix indices to the user equipment; and

repeating the steps of generating, computing and transmitting until receiving an acknowledgement from the user equipment that the data samples generated by the network node are correct.

15 . The apparatus of option ii of claim 9 , wherein the apparatus and the user equipment have agreed on using twin GAN, and the apparatus is caused to perform:

generating the desired set of data based on the suitable number of parts of the training data; and

training a machine learning model for a target goal using the desired set of training data.

Assignments (4)

Priority Claims (1)

Continuity (1)

References Cited (22)