Title:
MODEL TRAINING METHOD AND COMMUNICATION APPARATUS
Document Type and Number:
WIPO Patent Application WO/2024/017001
Kind Code:
A1
Abstract:
The present application relates to the field of communications. Disclosed are a model training method and a communication apparatus. The model training method comprises: receiving capability information which comes from a plurality of sub-node devices and is used for representing model training capabilities of the sub-node devices; sending, to each sub-node device among the plurality of sub-node devices, first indication information for indicating a first model parameter to be updated when the sub-node device trains a model, and the model; receiving updated first model parameters from the plurality of sub-node devices; fusing global model parameters of the plurality of sub-node devices by means of a preset fusion algorithm, so as to obtain a fused parameter; and sending the fused parameter to the plurality of sub-node devices. In this way, the problem of relatively low model training efficiency of existing federated-learning architectures is solved.
Inventors:
CHEN HONGZHI (CN)
LIU LIFU (CN)
SUN YAN (CN)
LIU LIFU (CN)
SUN YAN (CN)
Application Number:
PCT/CN2023/104254
Publication Date:
January 25, 2024
Filing Date:
June 29, 2023
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N20/00
Foreign References:
| CN111310932A | 2020-06-19 | |||
| CN114444708A | 2022-05-06 | |||
| CN114386570A | 2022-04-22 | |||
| CN112818207A | 2021-05-18 | |||
| US20210406782A1 | 2021-12-30 |
Attorney, Agent or Firm:
BEIJING ZBSD PATENT & TRADEMARK AGENT LTD. (CN)
Download PDF: