Title:
NEURAL NETWORK CONSTRUCTION METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2023/071793
Kind Code:
A1
Abstract:
The present application provides a neural network construction method. The method comprises: generating a parameter of a target neural network according to a parameter generation network, an input of the parameter generation network comprising information about a relative label of a neuron in the target neural network, the relative label of the neuron representing a relative position of the neuron in a first neural network layer, and the first neural network layer being a layer in the target neural network where the neuron is located; and constructing the target neural network according to the parameter of the target neural network. The parameter of the target neural network is generated by inputting the information about the relative label of the neuron in the target neural network, so that the parameter generation network has better generalization for constructing target neural networks of different dimensions.
Inventors:
WU JIAJUN (CN)
SUN CHENGJIAN (CN)
YANG CHENYANG (CN)
WANG JIAN (CN)
LI RONG (CN)
SUN CHENGJIAN (CN)
YANG CHENYANG (CN)
WANG JIAN (CN)
LI RONG (CN)
Application Number:
PCT/CN2022/124843
Publication Date:
May 04, 2023
Filing Date:
October 12, 2022
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/04
Foreign References:
| CN108229526A | 2018-06-29 | |||
| CN112418392A | 2021-02-26 | |||
| US20190385704A1 | 2019-12-19 | |||
| CN109063829A | 2018-12-21 | |||
| CN110163345A | 2019-08-23 |
Attorney, Agent or Firm:
LONGSUN LEAD IP LTD. (CN)
Download PDF:
Previous Patent: COVERED STENT AND COMBINED STENT
Next Patent: POLYCARBONATE ALLOY MATERIAL, PREPARATION THEREFOR, AND APPLICATION THEREOF
Next Patent: POLYCARBONATE ALLOY MATERIAL, PREPARATION THEREFOR, AND APPLICATION THEREOF