US 12,353,973 B2
Federated learning
Hui Lin, Beijing (CN); Jun Yang, Beijing (CN); Peng Fei Tian, Beijing (CN); and Yue Wang, Beijing (CN)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Sep. 17, 2021, as Appl. No. 17/478,196.
Prior Publication US 2023/0093067 A1, Mar. 23, 2023
Int. Cl. G06N 20/00 (2019.01); G06N 20/20 (2019.01)
CPC G06N 20/20 (2019.01) 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
maintaining a plurality of queues for storing local updates;
receiving, by at least one server computing device, the local updates epoch by epoch for a current global model from a plurality of client computing devices, each local update being determined by one of the plurality of client computing devices based on a local dataset on the client computing device, wherein retrieving the local updates from the plurality of queues comprises:
removing a local update from the queue and storing the removed local update in a multi-version (MV) store as a historical version of the local update;
responsive to the plurality of queues being empty in the next epoch, retrieving a historical version of the local update from the MV store for model training;
determining, by the at least one server computing device, a domain-specific aggregate of the local updates from each subset of the plurality of client computing devices; and
determining, by the at least one server computing device, an updated global model based on the domain-specific aggregate of the local updates for each subset of the plurality of client computing devices.