Title:
METHOD AND APPARATUS FOR TRAINING NEURAL NETWORK MODEL
Document Type and Number:
WIPO Patent Application WO/2021/051987
Kind Code:
A1
Abstract:
Provided is a method for training a neural network model in the field of artificial intelligence, comprising: acquiring a neural network model, first training data, and the class of the first training data, the neural network model being obtained by training on the basis of second training data (S401), the first training data comprising support data and query data, the support data comprising all or some of the data of each class in the first training data and the query data comprising all or some of the data of each class in the first training data; using the neural network model to perform feature extraction on the first training data to obtain the features of the first training data (S402); and, on the basis of the feature distance between the class-centric features of each class and the features of the query data, adjusting the parameters of some layers in the neural network model to obtain an adjusted neural network model (S403). A neural network model with good precision and generalisability is obtained by means of adjusting the parameters of some of the layers of the trained neural network model.
Inventors:
YU DEQUAN (CN)
WU JIHAO (CN)
JIA MINGBO (CN)
MA JIEYAN (CN)
WU JIHAO (CN)
JIA MINGBO (CN)
MA JIEYAN (CN)
Application Number:
PCT/CN2020/102594
Publication Date:
March 25, 2021
Filing Date:
July 17, 2020
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/08; G06N3/04
Foreign References:
| CN110175655A | 2019-08-27 | |||
| CN110175655A | 2019-08-27 | |||
| CN108875045A | 2018-11-23 | |||
| CN107480261A | 2017-12-15 | |||
| CN108898162A | 2018-11-27 | |||
| US20190147371A1 | 2019-05-16 |
Attorney, Agent or Firm:
LONGSUN LEAD IP LTD. (CN)
Download PDF: