Data processing method and apparatus, communication device, and storage medium
A data processing method includes determining, by a base station, a distribution characteristic of a local dataset of at least one user equipment (UE), and scheduling, by the base station, based on the distribution characteristic of the local dataset, a target UE from the at least one UE for participating in federated learning.
1 . A data processing method, comprising:
determining, by a base station, a distribution characteristic of a local dataset of at least one user equipment (UE);
scheduling, by the base station, based on the distribution characteristic, a target UE from the at least one UE for participating in federated learning; and
determining, based on statistical information of a distribution difference between the local dataset of the target UE and a global dataset of the base station, a weight coefficient of the target UE in the federated learning, wherein the global dataset is at least determined by the base station based on local datasets of a plurality of candidate UEs in the at least one UE associated with the base station,
wherein the statistical information of the distribution difference comprises: a probability distribution difference, and
the determining the weight coefficient of the target UE in the federated learning comprises:
determining the weight coefficient of the target UE based on a ratio of the probability distribution difference corresponding to the target UE to a sum of probability distribution differences corresponding to all UEs performing the federated learning.
2 . The method according to claim 1 , wherein the scheduling the target UE from the at least one UE for participating in the federated learning comprises:
obtaining statistical information of a distribution difference between the local dataset and a global dataset for the at least one UE; and
scheduling, based on the statistical information of the distribution difference, the target UE from the at least one UE for participating in the federated learning.
3 . The method according to claim 1 , further comprising:
obtaining capability information of the at least one UE; wherein
the scheduling the target UE from the at least one UE for participating in the federated learning comprises:
scheduling, based on the distribution characteristic and the capability information, the target UE from the at least one UE for participating in the federated learning.
4 . The method according to claim 3 , wherein the capability information of the at least one UE comprises at least one of:
computing capability information, indicating computing capability of the UE; or
communication status information, indicating at least one of a communication capability or a communication channel status of the UE.
5 . The method according to claim 4 , wherein the communication status information comprises: channel quality indication (CQI) information detected by the UE.
6 . The method according to claim 1 , further comprising:
receiving model information of a local model reported by the target UE for performing the federated learning; and
obtaining a global learning model by performing, based on the model information of the local model and the weight coefficient of the target UE, weighted averaging on the local model all UE performing the federated learning.
7 . The method according to claim 6 , further comprising:
terminating, in response to the global learning model satisfying a subscription requirement of operation administration and maintenance (OAM), reception of the model information.
8 . The method according to claim 7 , further comprising:
sending model information of the global learning model to the target UE in response to the global learning model not satisfying the subscription requirement of the OAM;
receiving model information of an updated local model updated by the target UE based on the global learning model; and
updating the global learning model based on the updated local model and the weight coefficient corresponding to the updated local model of the target UE.
9 . The method according to claim 1 , further comprising:
reporting model information of the global learning model and training data for training the global learning model to OAM;
receiving a model parameter determined by the OAM based on the model information of the global learning model, the training data and task data of the OAM; and
updating the global learning model based on the model parameter.
10 . The method according to claim 1 , further comprising:
determining, in response to detecting a handover occurs to the target UE connected to the base station, that the target UE exits the federated learning.
11 . A data processing method comprising:
receiving, by a user equipment (UE), scheduling information, wherein the scheduling information is sent by a base station based on a distribution characteristic of a local dataset of the UE for scheduling federated learning, wherein
the local dataset of the UE is used for the base station to determine, based on statistical information of a distribution difference between the local dataset of the UE and a global dataset of the base station, a weight coefficient of the UE in the federated learning, and the global dataset is at least determined by the base station based on local datasets of a plurality of candidate UEs in at least one UE associated with the base station,
wherein the statistical information of the distribution difference comprises: a probability distribution difference, and
the local dataset of the UE is used for the base station to determine the weight coefficient of the UE based on a ratio of the probability distribution difference corresponding to the UE to a sum of probability distribution differences corresponding to all UEs performing the federated learning.
12 . The method according to claim 11 , further comprising:
reporting capability information; and
receiving the scheduling information sent by the base station based on the distribution characteristic of the local dataset and the capability information.
13 . The method according to claim 12 , wherein the capability information comprises at least one of:
computing capability information, indicating computing capability of the UE; or
communication status information, indicating at least one of a communication capability or a communication channel status of the UE.
14 . The method according to claim 13 , wherein the communication status information comprises channel quality indication (CQI) information; and the method further comprises:
detecting the CQI information of a channel between the UE and the base station.
15 . The method according to claim 11 , further comprising:
reporting model information of a local model of the UE, wherein the local model is used for the base station to perform the federated learning based on the local model and the weight coefficient of the UE.
16 . The method according to claim 15 , further comprising:
generating the local dataset based on collected wireless network data;
generating a local training dataset by extracting data from the local dataset; and
obtaining the local model by performing model training using the local training dataset.
17 . The method according to claim 15 , further comprising:
receiving model information of a global learning model sent by the base station;
obtaining an updated local model by performing the federated learning based on the model information of the global learning model; and
reporting model information of the updated local model in response to the global learning model not satisfying a subscription requirement of operation administration and maintenance (OAM).
18 . A communication device, wherein the communication device at least comprises a processor and a memory used for storing executable instructions capable of running on the processor, wherein
the processor, through running the executable instructions, is configured to:
determine a distribution characteristic of a local dataset of at least one user equipment (UE);
schedule, based on the distribution characteristic, a target UE from the at least one UE for participating in federated learning; and
determine, based on statistical information of a distribution difference between the local dataset of the target UE and a global dataset of a base station, a weight coefficient of the target UE in the federated learning, wherein the global dataset is at least determined by the base station based on local datasets of a plurality of candidate UEs in the at least one UE associated with the base station,
wherein the statistical information of the distribution difference comprises a probability distribution difference, and
the processor is configured to determine the weight coefficient of the target UE based on a ratio of the probability distribution difference corresponding to the target UE to a sum of probability distribution differences corresponding to all UEs performing the federated learning; or
the processor, through running the executable instructions, is configured to:
receive scheduling information, wherein the scheduling information is sent by a base station based on a distribution characteristic of a local dataset of a UE for scheduling federated learning,
wherein the local dataset of the UE is used for the base station to determine, based on statistical information of a distribution difference between the local dataset of the UE and a global dataset of the base station, a weight coefficient of the UE in the federated learning, and the global dataset is at least determined by the base station based on local datasets of a plurality of candidate UEs in at least one UE associated with the base station,
wherein the statistical information of the distribution difference comprises a probability distribution difference, and
the local dataset of the UE is used for the base station to determine the weight coefficient of the UE based on a ratio of the probability distribution difference corresponding to the UE to a sum of probability distribution differences corresponding to all UEs performing the federated learning.