Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
LEARNING METHOD FOR ARTIFICIAL NEURAL NETWORK
Document Type and Number:
WIPO Patent Application WO/2020/005049
Kind Code:
A1
Abstract:
An embodiment of the present disclosure for solving the technical problem provides an information processing method performed by at least one processor. The information processing method may comprise the steps of: receiving first data by a processor; generating, by the processor, second data on the basis of the first data by using a first calculation model including multiple serially-connected sub-models, wherein the sub-models include at least one layer for changing a dimension; and generating, by the processor, a cost function on the basis of at least two pieces of data among the first data, the second data, and multiple pieces of sub-data generated by the multiple sub-models.

More Like This:
JP7403293Gaze detection using one or more neural networks
WO/2020/180518LEVERAGING LANGUAGE MODELS FOR GENERATING COMMONSENSE EXPLANATIONS
WO/2020/160252TASK-AWARE NEURAL NETWORK ARCHITECTURE SEARCH
Inventors:
SONG KIYOUNG (KR)
KANG HYEONG SHIN (KR)
Application Number:
PCT/KR2019/095019
Publication Date:
January 02, 2020
Filing Date:
May 23, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
SUALAB CO LTD (KR)
International Classes:
G06N3/08; G06N3/04
Foreign References:
KR101819857B12018-01-17
KR20180068292A2018-06-21
KR20180004046A2018-01-10
KR20180054709A2018-05-24
Other References:
"Autoencoder", 30 July 2016 (2016-07-30), Retrieved from the Internet
Attorney, Agent or Firm:
LEE, Dae Ho et al. (KR)
Download PDF:
View/Download PDF      PDF Help



 
Previous Patent: METHOD FOR VERIFYING DRONE INCLUDED IN INDUSTRIAL INTERNET OF THINGS SYSTEM, BY USING PETRI-NET MODE...

Next Patent: BACKPACK WITH INFLATABLE ELEMENT AND FASTENING ELEMENTS