Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
FEDERATED LEARNING METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2024/087573
Kind Code:
A1
Abstract:
A federated learning method and apparatus. The federated learning method comprises a center node sends a first model to at least one center edge device, receives at least one second model, and aggregates the at least one second model to obtain a fourth model. The at least one center edge device is in one-to-one correspondence to at least one edge device group. One second model is obtained by aggregating third models obtained by edge devices in at least one edge device group. One third model is a model obtained by learning the first model by at least one terminal device within an edge device collaborative coverage range on the basis of local data. A plurality of edge devices are divided into a plurality of edge device groups, and a center edge device in one edge device group sends a first model to each edge device in the edge device group, so that the number of edge devices interacting with a center node is reduced. In this way, the learning of models can be accelerated by increasing the number of edge devices participating in learning, thereby improving the federated learning efficiency.

Inventors:
SHAO YUNFENG (CN)
LI BINGSHUAI (CN)
LU JIAXUN (CN)
ZHENG ZHENZHE (CN)
WU FAN (CN)
HU DAHAI (CN)
Application Number:
PCT/CN2023/092742
Publication Date:
May 02, 2024
Filing Date:
May 08, 2023
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/098
Domestic Patent References:
WO2021247448A12021-12-09
WO2022175818A12022-08-25
Foreign References:
CN113469340A2021-10-01
CN113610303A2021-11-05
CN113255934A2021-08-13
CN113902021A2022-01-07
CN114363911A2022-04-15
CN114444708A2022-05-06
CN114828095A2022-07-29
CN114866545A2022-08-05
CN115174404A2022-10-11
Attorney, Agent or Firm:
TDIP & PARTNERS (CN)
Download PDF: