MULTI-TARGET ANALYSIS METHOD AND APPARATUS BASED ON REINFORCEMENT LEARNING FOR LEARNING UNDER-EXPLORED TARGET

A multi-target analysis method and apparatus are disclosed. The multi-target analysis apparatus comprises: an input and output unit for receiving an input of data, and outputting a result obtained by computing and processing the input data; a storage unit in which a program for performing a multi-ta...

Full description

Saved in:
Bibliographic Details
Main Authors KIM, Kibeom, ZHANG, Byoung-Tak, LEE, Min Whoo, HAN, Dong-Sig, LEE, Minsu, LEE, Hyundo
Format Patent
LanguageEnglish
French
Korean
Published 02.05.2024
Subjects
Online AccessGet full text

Cover

Loading…
Abstract A multi-target analysis method and apparatus are disclosed. The multi-target analysis apparatus comprises: an input and output unit for receiving an input of data, and outputting a result obtained by computing and processing the input data; a storage unit in which a program for performing a multi-target analysis method is stored; and a control unit which includes at least one processor, and which executes the program so as to analyze multiple targets received through the input and output unit, wherein the control unit collects a referring-expression-target pair in which a referring expression is matched with state information related to a target such that the target is specified through the referring expression, thereby generating a referring-expression-target set having a plurality of referring-expression-target pairs, and trains, with reference to the referring-expression-target set, a reinforcement learning-based learning model for receiving a referring expression related to a target and state information related to the target and outputting behavior information. L'invention concerne un procédé et un appareil d'analyse multi-cible. L'appareil d'analyse multi-cible comprend : une unité d'entrée et de sortie destinée à recevoir une entrée de données, et délivrer en sortie un résultat obtenu par calcul et traitement des données d'entrée ; une unité de stockage dans laquelle un programme pour effectuer un procédé d'analyse multi-cible est stocké ; et une unité de commande qui comprend au moins un processeur, et qui exécute le programme de façon à analyser de multiples cibles reçues par l'intermédiaire de l'unité d'entrée et de sortie, l'unité de commande collectant une paire cible d'expression de référence dans laquelle une expression de référence est mise en correspondance avec des informations d'état relatives à une cible de telle sorte que la cible est spécifiée par l'intermédiaire de l'expression de référence, générant ainsi un ensemble cible d'expression de référence ayant une pluralité de paires cibles d'expression de référence, et entraîne, en référence à l'ensemble cible d'expression de référence, un modèle d'apprentissage basé sur l'apprentissage de renforcement pour recevoir une expression de référence associée à une cible et des informations d'état relatives à la cible et délivrer des informations de comportement. 다중 타겟 분석 방법 및 장치를 제시한다. 다중 타겟 분석 장치는 데이터를 입력 받고, 이를 연산 처리한 결과를 출력하기 위한 입출력부, 다중 타겟 분석 방법을 수행하기 위한 프로그램이 저장되는 저장부, 및 적어도 하나의 프로세스를 포함하며, 상기 프로그램을 실행시킴으로써 상기 입출력부를 통해 수신된 다중 타겟을 분석하는 제어부를 포함하고, 상기 제어부는, 지시어를 통해 타겟이 특정되도록 상기 지시어 및 상기 타겟에 관한 상태 정보가 매칭된 지시어-타겟 쌍을 수집하여 복수의 지시어-타겟 쌍을 갖는 지시어-타겟 세트를 생성하고, 상기 지시어-타겟 세트를 참고하여, 상기 타겟에 관한 지시어 및 상기 타겟에 관한 상태 정보를 입력받아 행동 정보를 출력하는 강화 학습 기반의 학습 모델을 학습한다.
AbstractList A multi-target analysis method and apparatus are disclosed. The multi-target analysis apparatus comprises: an input and output unit for receiving an input of data, and outputting a result obtained by computing and processing the input data; a storage unit in which a program for performing a multi-target analysis method is stored; and a control unit which includes at least one processor, and which executes the program so as to analyze multiple targets received through the input and output unit, wherein the control unit collects a referring-expression-target pair in which a referring expression is matched with state information related to a target such that the target is specified through the referring expression, thereby generating a referring-expression-target set having a plurality of referring-expression-target pairs, and trains, with reference to the referring-expression-target set, a reinforcement learning-based learning model for receiving a referring expression related to a target and state information related to the target and outputting behavior information. L'invention concerne un procédé et un appareil d'analyse multi-cible. L'appareil d'analyse multi-cible comprend : une unité d'entrée et de sortie destinée à recevoir une entrée de données, et délivrer en sortie un résultat obtenu par calcul et traitement des données d'entrée ; une unité de stockage dans laquelle un programme pour effectuer un procédé d'analyse multi-cible est stocké ; et une unité de commande qui comprend au moins un processeur, et qui exécute le programme de façon à analyser de multiples cibles reçues par l'intermédiaire de l'unité d'entrée et de sortie, l'unité de commande collectant une paire cible d'expression de référence dans laquelle une expression de référence est mise en correspondance avec des informations d'état relatives à une cible de telle sorte que la cible est spécifiée par l'intermédiaire de l'expression de référence, générant ainsi un ensemble cible d'expression de référence ayant une pluralité de paires cibles d'expression de référence, et entraîne, en référence à l'ensemble cible d'expression de référence, un modèle d'apprentissage basé sur l'apprentissage de renforcement pour recevoir une expression de référence associée à une cible et des informations d'état relatives à la cible et délivrer des informations de comportement. 다중 타겟 분석 방법 및 장치를 제시한다. 다중 타겟 분석 장치는 데이터를 입력 받고, 이를 연산 처리한 결과를 출력하기 위한 입출력부, 다중 타겟 분석 방법을 수행하기 위한 프로그램이 저장되는 저장부, 및 적어도 하나의 프로세스를 포함하며, 상기 프로그램을 실행시킴으로써 상기 입출력부를 통해 수신된 다중 타겟을 분석하는 제어부를 포함하고, 상기 제어부는, 지시어를 통해 타겟이 특정되도록 상기 지시어 및 상기 타겟에 관한 상태 정보가 매칭된 지시어-타겟 쌍을 수집하여 복수의 지시어-타겟 쌍을 갖는 지시어-타겟 세트를 생성하고, 상기 지시어-타겟 세트를 참고하여, 상기 타겟에 관한 지시어 및 상기 타겟에 관한 상태 정보를 입력받아 행동 정보를 출력하는 강화 학습 기반의 학습 모델을 학습한다.
Author ZHANG, Byoung-Tak
KIM, Kibeom
LEE, Minsu
HAN, Dong-Sig
LEE, Min Whoo
LEE, Hyundo
Author_xml – fullname: KIM, Kibeom
– fullname: ZHANG, Byoung-Tak
– fullname: LEE, Min Whoo
– fullname: HAN, Dong-Sig
– fullname: LEE, Minsu
– fullname: LEE, Hyundo
BookMark eNqNjU8LgkAUxPdQh_59h4XOgpkGHV_61IX1reyuVCeR2E6lgn1_Wijo2mEYfsMMs2Szfujdgj2rRloRWNAFWg4E8mqE4RXaUmWeveoaNNjG8BMYzLgirlFQrnSKFZLlEkGToIL76AcNZagDvNRSab_6HKzZ_N49Jrf5-optc7RpGbhxaN00djfXu1d7VlEYxeExPCQJ7Pb_td6NAjnV
ContentType Patent
DBID EVB
DatabaseName esp@cenet
DatabaseTitleList
Database_xml – sequence: 1
  dbid: EVB
  name: esp@cenet
  url: http://worldwide.espacenet.com/singleLineSearch?locale=en_EP
  sourceTypes: Open Access Repository
DeliveryMethod fulltext_linktorsrc
Discipline Medicine
Chemistry
Sciences
Physics
DocumentTitleAlternate PROCÉDÉ ET APPAREIL D'ANALYSE MULTI-CIBLE BASÉS SUR L'APPRENTISSAGE PAR RENFORCEMENT POUR L'APPRENTISSAGE D'UNE CIBLE SOUS-EXPLORÉE
미탐색 타겟 학습을 위한 강화 학습 기반의 다중 타겟 분석 방법 및 장치
ExternalDocumentID WO2024090655A1
GroupedDBID EVB
ID FETCH-epo_espacenet_WO2024090655A13
IEDL.DBID EVB
IngestDate Fri Jul 19 12:56:12 EDT 2024
IsOpenAccess true
IsPeerReviewed false
IsScholarly false
Language English
French
Korean
LinkModel DirectLink
MergedId FETCHMERGED-epo_espacenet_WO2024090655A13
Notes Application Number: WO2022KR18709
OpenAccessLink https://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240502&DB=EPODOC&CC=WO&NR=2024090655A1
ParticipantIDs epo_espacenet_WO2024090655A1
PublicationCentury 2000
PublicationDate 20240502
PublicationDateYYYYMMDD 2024-05-02
PublicationDate_xml – month: 05
  year: 2024
  text: 20240502
  day: 02
PublicationDecade 2020
PublicationYear 2024
RelatedCompanies SEOUL NATIONAL UNIVERSITY R&DB FOUNDATION
RelatedCompanies_xml – name: SEOUL NATIONAL UNIVERSITY R&DB FOUNDATION
Score 3.512619
Snippet A multi-target analysis method and apparatus are disclosed. The multi-target analysis apparatus comprises: an input and output unit for receiving an input of...
SourceID epo
SourceType Open Access Repository
SubjectTerms CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
PHYSICS
Title MULTI-TARGET ANALYSIS METHOD AND APPARATUS BASED ON REINFORCEMENT LEARNING FOR LEARNING UNDER-EXPLORED TARGET
URI https://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240502&DB=EPODOC&locale=&CC=WO&NR=2024090655A1
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwfV3dT8IwEL8Q_HxT1PiBpolmb4tzX24PxJStwMzYlm0oPpExSmJUIDLjv--tgPDEQx_umrZpk-v11_7uCnA3HlqGrnFFVjJTkXUjt8tHQkXORmUkpjW2H3LBtgjMTk9_7hv9CnyuYmFEntBfkRwRLSpHey_Efj1bX2K5gls5vx--o2r61EobrrREx-ieDEWV3GaDRaEbOpLjIG6TgnhRZ6O_NShipR08SD-WBDD20izjUmabTqV1BLsR9jcpjqHyMa3BgbP6e60G-93lk3cN9gRHM5-jcmmH8xP46vb81JNTGrdZSmhA_bfES0iXpZ3QRRlLFNGYpr2ENGnCXBIGJGZegJjPEfn7ic9oHHhBm6BqLYiYA5n1Iz-MsdVigFO4bbHU6cg4g8H_gg1ew83pamdQnUwn_BwI1xCtmOqIcyvXzQxRtWVyTTM4z1TTHisXUN_W0-X26is4LEVBB1TrUC2-f_g1uuxieCNW-g-49ZBm
link.rule.ids 230,309,783,888,25578,76884
linkProvider European Patent Office
linkToHtml http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwfV3dT8IwEG8IfuCbosYP1CaavS3OjS3bAzFjK2y6dctWFJ_IGCUxKhCZ8d_3VkB44qEPd5e26SXX66-9uyJ0Nx6aelPjiqxkhiI39dwqHwkVORuVmZjm2HrIRbQFNbxe86mv9yvoc5ULI-qE_oriiGBROdh7Ifbr2foSyxWxlfP74Tuwpo8d1nKlJToG96QrquS2WySO3MiRHAdwm0SThcwCf6vbgJV24JBtlpX2yUu7zEuZbTqVziHajWG8SXGEKh_TOqo5q7_X6mg_XD5519GeiNHM58Bc2uH8GH2FvYD5MrOTLmHYpnbwlvopDgnzIhdoaHFsJzbrpbhtp8TFEcUJ8SlgPkfU78cBsRPq0y4G1poQOQcy6cdBlECvxQQn6LZDmOPJsILBv8IGr9HmcrVTVJ1MJ_wMYa4BWjHUEedm3jQyQNWmwTVN5zxTDWusnKPGtpEutotvUM1jYTAIfPp8iQ5KkQgNVBuoWnz_8Ctw38XwWmj9D6gEk1Y
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Apatent&rft.title=MULTI-TARGET+ANALYSIS+METHOD+AND+APPARATUS+BASED+ON+REINFORCEMENT+LEARNING+FOR+LEARNING+UNDER-EXPLORED+TARGET&rft.inventor=KIM%2C+Kibeom&rft.inventor=ZHANG%2C+Byoung-Tak&rft.inventor=LEE%2C+Min+Whoo&rft.inventor=HAN%2C+Dong-Sig&rft.inventor=LEE%2C+Minsu&rft.inventor=LEE%2C+Hyundo&rft.date=2024-05-02&rft.externalDBID=A1&rft.externalDocID=WO2024090655A1