Title:
MULTITASK LEARNING METHOD AND DEVICE
Document Type and Number:
WIPO Patent Application WO/2021/259305
Kind Code:
A1
Abstract:
A multitask learning method and device. The method comprises: inputting training data in an information source of a plurality of tasks to a coding network for coding, to obtain coded training data, the coded training data comprising feature information of the plurality of tasks, the feature information of the plurality of tasks comprising private feature information of each task and/or common feature information between each task and other tasks, and the information source of the plurality of tasks comprising an annotation result of the training data (S1301); inputting the coded training data to a decoding network for decoding to obtain a prediction result of the training data (S1302); adjusting a first neural network parameter of the coding network according to the feature information of the plurality of tasks and a constraint condition (S1303); and adjusting a second neural network parameter of the decoding network according to an error between the annotation result of the training data and the prediction result of the training data (S1304). By means of the method, the application of multitask learning in wireless communication is achieved.
Inventors:
KONG CHUILI (CN)
ZHANG GONGZHENG (CN)
LI RONG (CN)
ZHANG GONGZHENG (CN)
LI RONG (CN)
Application Number:
PCT/CN2021/101723
Publication Date:
December 30, 2021
Filing Date:
June 23, 2021
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/04
Foreign References:
| CN110097528A | 2019-08-06 | |||
| CN110574049A | 2019-12-13 | |||
| CN111247532A | 2020-06-05 | |||
| CN105404877A | 2016-03-16 | |||
| US20190244108A1 | 2019-08-08 |
Other References:
See also references of EP 4156028A4
Attorney, Agent or Firm:
TDIP & PARTNERS (CN)
Download PDF: