Method and apparatus to federated machine learning in wireless communication system
The disclosure relates to a 5th generation (5G) or 6th generation (6G) communication system for supporting a higher data transmission rate. A method of performing federated learning by a first network data analytic function (NWDAF) is provided. The method includes determining whether to perform a federated learning based on an analytics request message from a network function (NF), receiving, from a network repository function (NRF), information regarding a second NWDAF capable of participating in the federated learning, generating a federated learning group, transmitting, to the second NWDAF, a federated learning join request message based on the information regarding the second NWDAF, and receiving, from the second NWDAF, a federated learning join response message including information regarding whether to participate in the federated learning group.
1 . A method of performing federated learning (FL) by a network repository function (NRF), the method comprising:
receiving, from a central network data analytic function (NWDAF), a first registration request message including information on a FL capability type;
transmitting, to the central NWDAF, a first registration response message;
receiving, from a plurality of local NWDAFs, a plurality of second registration request messages;
transmitting, to the plurality of local NWDAFs, a plurality of second registration response message;
receiving, from the central NWDAF, a discovery request message; and
transmitting, to the central NWDAF, a discovery response message including information about the plurality of the local NWDAFs,
wherein at least one NWDAF for a federated learning group is determined among the plurality of the local NWDAFs, based on information on a network function (NF) of a data source, and information, included in respective federated learning join response messages received from the plurality of the local NWDAFs, indicating whether to participate in the federated learning group.
2 . The method of claim 1 , wherein the at least one NWDAF for the federated learning group is further determined based on at least one of information on FL capability type and the information on a service area.
3 . The method of claim 1 ,
wherein the central NWDAF requests the at least one NWDAF for the federated learning group to train a machine learning model and reports machine learning model information.
4 . The method of claim 1 , wherein the central NWDAF generates a global machine learning model based on the machine learning model information obtained from the at least one NWDAF for the federated learning group and sends the global machine learning model to the at least one NWDAF for the federated learning group.
5 . A method of performing federated learning (FL) by a central network data analytic function (NWDAF), the method comprising:
transmitting, to a network repository function (NRF), a registration request message including information on FL capability type;
receiving, from the NRF, a registration response message;
transmitting, to the NRF, a discovery request message;
receiving, from the NRF, a discovery response message including information about a plurality of local NWDAFs;
transmitting, to a plurality of local NWDAFs, respective federated learning join request messages; and
receiving, from the plurality of local NWDAFs, respective federated learning join response messages including information indicating whether to participate in a federated learning group,
wherein at least one NWDAF for a federated learning group is determined among the plurality of the local NWDAFs, based on information on a network function (NF) of a data source, and information, included in respective federated learning join response messages received from the plurality of the local NWDAFs, indicating whether to participate in the federated learning group.
6 . The method of claim 5 , wherein the at least one NWDAF for the federated learning group is further determined based on at least one of information on FL capability type and information on a service area.
7 . The method of claim 5 ,
wherein the central NWDAF generates a global machine learning model based on the machine learning model information obtained from the at least one NWDAF for the federated learning group and sends the global machine learning model to the at least one NWDAF for the federated learning group.
8 . The method of claim 5 , wherein the transmitting the discovery request message comprises transmitting the discovery request message, based on whether the federated learning is required, which is determined based on an analytic ID.
9 . A network repository function (NRF) of performing federated learning (FL), the NRF comprising:
a transceiver; and
at least one processor coupled with the transceiver and configured to:
receive, from a central network data analytic function (NWDAF), a first registration request message including information on a FL capability type;
transmit, to the central NWDAF, a first registration response message;
receive, from a plurality of local NWDAFs, a plurality of second registration request messages;
transmit to the plurality of local NWDAFs, a plurality of second registration response message;
receive, from the central NWDAF, a discovery request message; and
transmit, to the central NWDAF, a discovery response message including information about the plurality of the local NWDAFs,
wherein at least one NWDAF for a federated learning group is determined among the plurality of the local NWDAFs, based on information on a network function (NF) of a data source, and information, included in respective federated learning join response messages received from the plurality of the local NWDAFs, indicating whether to participate in the federated learning group.
10 . The NRF of claim 9 , wherein the at least one NWDAF for the federated learning group is further determined based on at least one of information on FL capability type and information on a service area.
11 . The NRF of claim 9 ,
wherein the central NWDAF requests the at least one NWDAF for the federated learning group to train a machine learning model and reports machine learning model information.
12 . The NRF of claim 9 , wherein the central NWDAF generates a global machine learning model based on the machine learning model information obtained from the at least one NWDAF for the federated learning group and sends the global machine learning model to the at least one NWDAF for the federated learning group.
13 . A central network data analytic function (NWDAF) of performing federated learning (FL), the central NWDAF comprising:
a transceiver; and
at least one processor coupled with the transceiver and configured to:
transmit, to a network repository function (NRF), a registration request message including information on FL capability type,
receive, from the NRF, a registration response message,
transmit, to the NRF, a discovery request message,
receive, from the NRF, a discovery response message including information about a plurality of local NWDAFs,
transmit, to a plurality of local NWDAFs, respective federated learning join request messages; and
receive, from the plurality of local NWDAFs, respective federated learning join response messages including information indicating whether to participate in a federated learning group,
wherein at least one NWDAF for a federated learning group is determined among the plurality of the local NWDAFs, based on information on a network function (NF) of a data source, and information, included in respective federated learning join response messages received from the plurality of the local NWDAFs, indicating whether to participate in the federated learning group.
14 . The central NWDAF of claim 13 , wherein the at least one NWDAF for the federated learning group is further determined based on at least one of information on FL capability type and information on a service area.
15 . The central NWDAF of claim 13 ,
wherein the central NWDAF generates a global machine learning model based on the machine learning model information obtained from the at least one NWDAF for the federated learning group and sends the global machine learning model to the at least one NWDAF for the federated learning group.
16 . The central NWDAF of claim 13 , wherein the at least one processor is further configured to transmit the discovery request message, based on whether the federated learning is required, which is determined based on an analytic ID.