Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
USER DEVICE SELECTION METHOD AND APPARATUS, AND CHIP AND MODULE DEVICE
Document Type and Number:
WIPO Patent Application WO/2023/143082
Kind Code:
A1
Abstract:
Disclosed in the present application are a user device selection method and apparatus, and a chip and a module device. The method comprises: sending a first message to a first candidate user device group; receiving a second message from a second candidate user device group, wherein the first candidate user device group comprises the second candidate user device group, the first message indicates demand information for federated learning, and the second message indicates whether a user device satisfies the demand information and/or indicates measurement information of the corresponding user device; or the first message is used for requesting the measurement information, and the second message comprises the measurement information; and the demand information for federated learning comprises duration demand information of federated learning, and/or demand information of federated learning for the stability of the user device; and determining, on the basis of the second message, a target user device group that participates in federated learning, wherein each target user device in the target user device group satisfies the demand information. By means of the method, the stability of a user device participating in federated learning can be improved, thereby reducing a federated learning interruption probability.

Inventors:
CHEN XIAOYU (CN)
HAN LIFENG (CN)
Application Number:
PCT/CN2023/071729
Publication Date:
August 03, 2023
Filing Date:
January 10, 2023
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
SPREADTRUM COMM SHANGHAI CO (CN)
International Classes:
G06N20/00; H04W24/10
Foreign References:
CN111538598A2020-08-14
CN110598870A2019-12-20
CN111428885A2020-07-17
CN113255924A2021-08-13
US20210398026A12021-12-23
Attorney, Agent or Firm:
SCIHEAD IP LAW FIRM (CN)
Download PDF: