Method, system, and non-transitory computer-readable record medium for providing multiple models of federated learning using personalization
Disclosed is a method, system, and non-transitory computer-readable record medium for providing a multi-model through federated learning using personalization. The method includes classifying users into a plurality of groups; and generating a prediction model for a service as a multi-model through federated learning for each of the plurality of groups.
1. A method performed by a computer system comprising at least one processor configured to execute computer-readable instructions comprised in a memory, the method comprising:
by the at least one processor,
distributing an initial model generated by the computer system to electronic devices of users;
classifying the users into a plurality of groups, at least one first user among the users being classified into at least two groups such that the first user belongs to all of the two groups; and
generating a prediction model for a service as a multi-model through federated learning for each of the plurality of groups,
wherein the generating of the prediction model for the service comprises
generating a corresponding group model, for each of the plurality of groups, by collecting, into a single model of the multi-model, models trained by the electronic devices of the users belonging to a corresponding group through the federated learning,
distributing the corresponding group model to the electronic devices of the users belonging to the corresponding group, each of at least two corresponding group models corresponding to the two groups being distributed to an electronic device of the first user, the at least two corresponding group models including a first group model related to a first chatroom where the first user participates and a second group model related to a second chatroom where the first user participates, the first group model being trained based on data associated with the first chatroom in the electronic device of the first user and first update data associated with the first chatroom for the first group model being received from the electronic device of the first user, and the second group model trained being based on data associated with the second chatroom in the electronic device of the first user and second update data associated with the second chatroom for the second group model being received from the electronic device of the first user, and
improving the corresponding group model based on the update data about the corresponding group model received from the electronic devices of the users belonging to the corresponding group,
wherein the first group model is improved based on a first update data set of the update data including the first update data and the second group model is improved based on a second update data set of the update data including the second update data.
2. The method of claim 1 , wherein the classifying of the users comprises grouping the users based on at least one piece of information collectable by the computer system in association with the service.
3. The method of claim 1 , wherein the classifying of the users comprises grouping the users based on at least one of a user profile or domain knowledge of the service.
4. The method of claim 1 , wherein the classifying of the users comprises grouping the users based on collaborative filtering.
5. The method of claim 1 , wherein
the service is a content recommendation service, and
the generating of the prediction model for the service comprises generating a content recommendation model for each of the plurality of groups as a learning result model using in-device data related to content in an electronic device of each of the users belonging to a corresponding group through the federated learning for each group of the plurality of groups.
6. The method of claim 5 , wherein the classifying of the users comprises grouping the users based on at least one piece of information collectable by the computer system in association with the content recommendation service.
7. The method of claim 1 , wherein the federated learning includes learning the multi-model of the prediction model based on a plurality of the single model.
8. The method of claim 1 , wherein the method further comprises:
by the at least one processor,
providing the improved corresponding group model to the users belonging to the corresponding group in association with a provision of the service for the users belonging to the corresponding group,
classifying at least one of other users not participating in the federated learning into the corresponding group, and
providing the improved corresponding group model to the at least one of other users belonging to the corresponding group in association with a provision of the service for the at least one of other users belonging to the corresponding group.
9. A non-transitory computer-readable record medium storing instructions that, when executed by a processor, cause the processor to perform the method of claim 1 .
10. A computer system comprising:
at least one processor configured to execute computer-readable instructions comprised in a memory,
wherein the at least one processor is configured to
provide an initial model generated by the computer system to electronic devices of users;
classify the users into a plurality of groups, at least one first user among the users being classified into at least two groups such that the first user belongs to all of the two groups; and
generate a prediction model for a service as a multi-model through federated learning for each of the plurality of groups and to provide the generated prediction model,
wherein, for generating the prediction model, the at least one processor is configured to generate a corresponding group model, for each of the plurality of groups, by collecting, into a single model of the multi-model, models trained by the electronic devices of the users belonging to a corresponding group through the federated learning,
distribute the corresponding group model to the electronic devices of the users belonging to the corresponding group, each of at least two corresponding group models corresponding to the two groups being distributed to an electronic device of the first user and said each of the two corresponding group models being individually trained based on data corresponding to a respective group model of the two corresponding group models in the electronic device of the first user to generate update data about said each of the two corresponding group models, the at least two corresponding group models including a first group model related to a first chatroom where the first user participates and a second group model related to a second chatroom where the first user participates, the first group model being trained based on data associated with the first chatroom in the electronic device of the first user and first update data associated with the first chatroom for the first group model being received from the electronic device of the first user, and the second group model being trained based on data associated with the second chatroom in the electronic device of the first user and second update data associated with the second chatroom for the second group model being received from the electronic device of the first user and
improve the corresponding group model based on update data about the corresponding group model received from the electronic devices of the users belonging to the corresponding group,
wherein the first group model is improved based on a first update data set of the update data including the first update data and the second group model is improved based on a second update data set of the update data including the second update data.
11. The computer system of claim 10 , wherein the at least one processor is configured to group the users based on at least one piece of information collectable by the computer system in association with the service.
12. The computer system of claim 10 , wherein the at least one processor is configured to group the users based on at least one of a user profile or domain knowledge of the service.
13. The computer system of claim 10 , wherein
the service is a content recommendation service, and
the at least one processor is configured to generate a content recommendation model for each of the plurality of groups as a learning result model using in-device data related to the content in an electronic device of each user belonging to a corresponding group through federated learning for each of the plurality of groups.
14. The computer system of claim 13 , wherein the at least one processor is configured to group the users based on at least one piece of information collectable by the computer system in association with the content recommendation service.
15. The computer system of claim 10 , the at least one processor is further configured to
provide the improved corresponding group model to the users belonging to the corresponding group in association with a provision of the service for the users belonging to the corresponding group,
classify at least one of other users not participating in the federated learning into the corresponding group, and
provide the improved corresponding group model to the at least one of other users belonging to the corresponding group in association with a provision of the service for the at least one of other users belonging to the corresponding group.