Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020091139 - EFFECTIVE NETWORK COMPRESSION USING SIMULATION-GUIDED ITERATIVE PRUNING

Publication Number WO/2020/091139
Publication Date 07.05.2020
International Application No. PCT/KR2018/015831
International Filing Date 13.12.2018
IPC
G06N 3/08 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
G06N 3/04 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
Applicants
  • 주식회사 노타 NOTA, INC. [KR]/[KR]
Inventors
  • 정대웅 JEONG, Dae-Woong
  • 김재헌 KIM, Jaehun
  • 김영석 KIM, Young Seok
  • 채명수 CHAE, Myungsu
Agents
  • 양성보 YANG, Sungbo
Priority Data
10-2018-013165531.10.2018KR
10-2018-015675007.12.2018KR
Publication Language Korean (KO)
Filing Language Korean (KO)
Designated States
Title
(EN) EFFECTIVE NETWORK COMPRESSION USING SIMULATION-GUIDED ITERATIVE PRUNING
(FR) COMPRESSION DE RÉSEAU EFFICACE À L'AIDE D'UN ÉLAGAGE ITÉRATIF GUIDÉ PAR SIMULATION
(KO) 시뮬레이션-가이드된 반복적 프루닝을 사용하는 효율적인 네트워크 압축
Abstract
(EN)
The effective network compression using simulation-guided iterative pruning, according to various embodiments, can be configured so that, by means of an electronic device, a first neural network is pruned on the basis of a threshold value, a second neural network is generated, a gradient for each weighted value of the second neural network is calculated, and a third neural network is acquired by applying the gradient to the first neural network.
(FR)
L'invention concerne une compression de réseau efficace à l'aide d'un élagage itératif guidé par simulation qui, selon divers modes de réalisation, peut être configurée de telle sorte que, au moyen d'un dispositif électronique, un premier réseau neuronal est élagué sur la base d'une valeur de seuil, un deuxième réseau neuronal est généré, un gradient pour chaque valeur pondérée du deuxième réseau neuronal est calculé, et un troisième réseau neuronal est acquis par application du gradient au premier réseau neuronal.
(KO)
다양한 실시예들에 따른 시뮬레이션-가이드된 반복적 프루닝을 사용하는 효율적인 네트워크 압축은, 전자 장치에 의해, 제 1 뉴럴 네트워크를 임계 값에 기반하여 프루닝하여, 제 2 뉴럴 네트워크를 생성하고, 제 2 뉴럴 네트워크의 각 가중치에 대한 그라디언트를 계산하고, 그라디언트를 제 1 뉴럴 네트워크에 적용하여, 제 3 뉴럴 네트워크를 획득하도록 구성될 수 있다.
Latest bibliographic data on file with the International Bureau