Some content of this application is unavailable at the moment.
If this situation persist, please contact us atFeedback&Contact
1. (WO2019050297) NEURAL NETWORK LEARNING METHOD AND DEVICE
Latest bibliographic data on file with the International Bureau    Submit observation

Pub. No.: WO/2019/050297 International Application No.: PCT/KR2018/010421
Publication Date: 14.03.2019 International Filing Date: 06.09.2018
IPC:
G06N 3/08 (2006.01) ,G06N 3/04 (2006.01)
G PHYSICS
06
COMPUTING; CALCULATING; COUNTING
N
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3
Computer systems based on biological models
02
using neural network models
08
Learning methods
G PHYSICS
06
COMPUTING; CALCULATING; COUNTING
N
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3
Computer systems based on biological models
02
using neural network models
04
Architecture, e.g. interconnection topology
Applicants:
삼성전자 주식회사 SAMSUNG ELECTRONICS CO., LTD. [KR/KR]; 경기도 수원시 영통구 삼성로 129 129, Samsung-ro, Yeongtong-gu, Suwon-si, Gyeonggi-do 16677, KR
한국과학기술원 KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY [KR/KR]; 대전시 유성구 대학로 291 291, Daehak-ro, Yuseong-gu, Daejeon 34141, KR
Inventors:
김준모 KIM, Jun-mo; KR
이예강 LEE, Ye-gang; KR
김병주 KIM, Byung-ju; KR
이시행 LEE, Si-haeng; KR
박민석 PARK, Min-seok; KR
안평환 AHN, Pyung-whan; KR
이재영 LEE, Jae-young; KR
Agent:
리앤목 특허법인 Y.P.LEE, MOCK & PARTNERS; 서울시 강남구 언주로30길 13 대림아크로텔 12층 12F Daelim Acrotel 13 Eonju-ro 30-gil, Gangnam-gu, Seoul 06292, KR
Priority Data:
10-2017-011546408.09.2017KR
Title (EN) NEURAL NETWORK LEARNING METHOD AND DEVICE
(FR) PROCÉDÉ ET DISPOSITIF D'APPRENTISSAGE DE RÉSEAU NEURONAL
(KO) 뉴럴 네트워크 학습 방법 및 장치
Abstract:
(EN) The present disclosure relates to: an artificial intelligence (AI) system for mimicking functions of the human brain such as cognition, judgment, etc. by using a machine learning algorithm such as deep learning; and an application thereof. Especially, the present disclosure relates to a neural network learning method based on the AI system and the application thereof. The method enables: from a filter comprising weighted value information of at least one hidden layer included in a learned network model, extracting properties of input data according to the weighted value information, corresponding to a specific part, of the filter by using a masking filter exhibiting an effective value at the specific part; on the basis of the extracted properties of the input data, comparing target data and output data acquired from the learned network model; and on the basis of the comparison result, updating the size of the specific part at which the masking filter exhibits the effective value.
(FR) La présente invention concerne : un système d'intelligence artificielle (AI) destiné à imiter des fonctions du cerveau humain telles que la cognition, la détermination, etc. à l'aide d'un algorithme d'apprentissage automatique tel qu'un apprentissage profond, ainsi qu'une application de celui-ci. En particulier, la présente invention concerne un procédé d'apprentissage de réseau neuronal basé sur le système AI et son application. Le procédé permet : à partir d'un filtre comprenant des informations de valeur pondérée d'au moins une couche cachée comprise dans un modèle de réseau appris, d'extraire des propriétés de données d'entrée selon les informations de valeur pondérée, correspondant à une partie spécifique, du filtre à l'aide d'un filtre de masquage présentant une valeur efficace au niveau de la partie spécifique; sur la base des propriétés extraites des données d'entrée, de comparer des données cibles et des données de sortie acquises à partir du modèle de réseau appris; et sur la base du résultat de la comparaison, de mettre à jour la taille de la partie spécifique au niveau de laquelle le filtre de masquage présente la valeur efficace.
(KO) 본 개시는 딥러닝 등의 기계 학습 알고리즘을 활용하여 인간 두뇌의 인지, 판단 등의 기능을 모사하는 인공지능(AI) 시스템 및 그 응용에 관련된 것이다. 특히, 본 개시는 인공지능 시스템 및 그 응용에 따른 뉴럴 네트워크의 학습 방법으로, 학습 네트워크 모델에 포함된 적어도 하나의 히든 레이어의 가중치 정보로 구성된 필터로부터, 특정 부분에 유효값을 갖는 마스킹 필터를 이용하여, 특정 부분에 대응되는 필터의 가중치 정보에 따라 입력 데이터의 특징을 추출하고, 추출된 입력 데이터의 특징을 기초로 학습 네트워크 모델로부터 획득된 출력 데이터와 타겟 데이터를 비교하며, 비교 결과에 기초하여, 마스킹 필터에서 유효값을 갖는 특정 부분의 크기를 업데이트 할 수 있다.
front page image
Designated States: AE, AG, AL, AM, AO, AT, AU, AZ, BA, BB, BG, BH, BN, BR, BW, BY, BZ, CA, CH, CL, CN, CO, CR, CU, CZ, DE, DJ, DK, DM, DO, DZ, EC, EE, EG, ES, FI, GB, GD, GE, GH, GM, GT, HN, HR, HU, ID, IL, IN, IR, IS, JO, JP, KE, KG, KH, KN, KP, KW, KZ, LA, LC, LK, LR, LS, LU, LY, MA, MD, ME, MG, MK, MN, MW, MX, MY, MZ, NA, NG, NI, NO, NZ, OM, PA, PE, PG, PH, PL, PT, QA, RO, RS, RU, RW, SA, SC, SD, SE, SG, SK, SL, SM, ST, SV, SY, TH, TJ, TM, TN, TR, TT, TZ, UA, UG, US, UZ, VC, VN, ZA, ZM, ZW
African Regional Intellectual Property Organization (ARIPO) (BW, GH, GM, KE, LR, LS, MW, MZ, NA, RW, SD, SL, ST, SZ, TZ, UG, ZM, ZW)
Eurasian Patent Office (AM, AZ, BY, KG, KZ, RU, TJ, TM)
European Patent Office (EPO) (AL, AT, BE, BG, CH, CY, CZ, DE, DK, EE, ES, FI, FR, GB, GR, HR, HU, IE, IS, IT, LT, LU, LV, MC, MK, MT, NL, NO, PL, PT, RO, RS, SE, SI, SK, SM, TR)
African Intellectual Property Organization (BF, BJ, CF, CG, CI, CM, GA, GN, GQ, GW, KM, ML, MR, NE, SN, TD, TG)
Publication Language: Korean (KO)
Filing Language: Korean (KO)