公开(公告)号：US20220012601A1

公开(公告)日：2022-01-13

申请号：US17484886

申请日：2021-09-24

Applicant: Huawei Technologies Co., Ltd.

Inventor： Muhammad AMAD-UD-DIN ,  Adrian FLANAGAN ,  Kuan Eeik TAN ,  Elena IVANNIKOVA ,  Qiang FU

IPC: G06N5/02 ,  G06N5/00

Abstract: A Federated learning server and a method are provided. The Federated learning server is configured to aggregate a plurality of received model updates to update a master machine learning model. Once a pre-defined threshold or interval for received model updates is reached, a set of current hyper-parameter values and corresponding validation set performance metrics obtained from the updated master machine learning model are sent to a hyper-parameter optimization model. The optimization model infers the next set of optimal hyper-parameters using pairwise history of hyper-parameter values and the corresponding performance metrics. The inferred hyper-parameter values are sent to the Federated Learning server which updates the master machine learning model with the updated set of hyper-parameter values and redistributes the updated master machine learning model with the updated set of hyper-parameter values. According to the application, hyper-parameter optimization in a Federated learning mode can be realized to provide accurate personalized recommendations.