Title:
GENERATIVE ADVERSARIAL NEURAL NETWORK TRAINING METHOD AND SYSTEM
Document Type and Number:
WIPO Patent Application WO/2021/174935
Kind Code:
A1
Abstract:
A generative adversarial neural network training method, comprising the following steps: establishing an initial discrimination neural network and an initial generation neural network so as to form an initial generative adversarial neural network (S100); initializing a parameter of the initial generative adversarial neural network and a boundary vector (S102); acquiring a real sample set and a random variable set, and inputting the random variable set into the initial generation neural network so as to generate a false sample set (S104); inputting the real sample set and the false sample set into the initial discrimination neural network so as to obtain a first discrimination output and a second discrimination output (S106); performing a calculation according to a preset discriminant loss function to obtain a discriminant loss value (S108); performing a calculation according to a preset generation loss function to obtain a generation loss value (S110); and updating the parameter of the initial generative adversarial neural network according to the discriminant loss value and the generation loss value so as to obtain a target generative adversarial neural network (S112). By means of the method, the training speed and stability of a generative adversarial neural network can be improved.
Inventors:
CHEN ZHUOJUN (CN)
LU JIN (CN)
CHEN BIN (CN)
SONG CHEN (CN)
LU JIN (CN)
CHEN BIN (CN)
SONG CHEN (CN)
Application Number:
PCT/CN2020/134889
Publication Date:
September 10, 2021
Filing Date:
December 09, 2020
Export Citation:
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G06N3/04
Foreign References:
CN111445007A | 2020-07-24 | |||
CN107180392A | 2017-09-19 | |||
CN108960278A | 2018-12-07 | |||
CN107563995A | 2018-01-09 | |||
US20190220388A1 | 2019-07-18 |
Attorney, Agent or Firm:
SCIHEAD IP LAW FIRM (CN)
Download PDF: