ELECTRONIC APPARATUS AND CONTROL METHOD THEREFOR

Provided are an electronic apparatus and a control method therefor. The electronic apparatus comprises: a memory for storing at least one instruction; and at least one processor. The at least one processor executes the at least one instruction to: obtain a first quantization matrix expressed as a pr...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: JEON, Yongkweon, PARK, Kyungphil, KIM, Hoyoung, LEE, Chungman
Format: Patent
Sprache:eng ; fre ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator JEON, Yongkweon
PARK, Kyungphil
KIM, Hoyoung
LEE, Chungman
description Provided are an electronic apparatus and a control method therefor. The electronic apparatus comprises: a memory for storing at least one instruction; and at least one processor. The at least one processor executes the at least one instruction to: obtain a first quantization matrix expressed as a product of a first vector and a second vector for each of a plurality of layers on the basis of a weight for each of the plurality of layers included in a neural network model; obtain a second quantization matrix for each of the plurality of layers by multiplying each of first vectors of first quantization matrices corresponding to the plurality of layers by a second vector of a first quantization matrix corresponding to a previous layer; and quantize the neural network model by using the second quantization matrix. L'invention concerne un appareil électronique et son procédé de commande. L'appareil électronique comprend : une mémoire destinée à stocker au moins une instruction ; et au moins un processeur. Ledit au moins un processeur exécute ladite au moins une instruction pour : obtenir une première matrice de quantification exprimée en tant que produit d'un premier vecteur et d'un second vecteur pour chacune d'une pluralité de couches sur la base d'un poids pour chacune de la pluralité de couches comprises dans un modèle de réseau de neurones artificiels ; obtenir une seconde matrice de quantification pour chacune de la pluralité de couches par multiplication de chacun des premiers vecteurs de premières matrices de quantification correspondant à la pluralité de couches par un second vecteur d'une première matrice de quantification correspondant à une couche précédente ; et quantifier le modèle de réseau de neurones artificiels à l'aide de la seconde matrice de quantification. 전자 장치 및 이의 제어 방법이 제공된다. 본 전자 장치는 적어도 하나의 인스트럭션을 저장하는 메모리 및 적어도 하나의 프로세서를 포함한다. 적어도 하나의 프로세서는, 적어도 하나의 인스트럭션을 실행함으로써, 신경망 모델에 포함된 복수의 레이어 별 가중치에 기초하여 복수의 레이어 별로 제1 벡터 및 제2 벡터의 곱으로 표현된 제1 양자화 행렬을 획득하며, 복수의 레이어에 대응되는 제1 양자화 행렬들의 제1 벡터 각각을 이전 레이어에 대응되는 제1 양자화 행렬의 제2 벡터와 곱하여 복수의 레이어 별로 제2 양자화 행렬을 획득하며, 제2 양자화 행렬을 이용하여 신경망 모델을 양자화한다.
format Patent
fullrecord <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_WO2024258068A1</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>WO2024258068A1</sourcerecordid><originalsourceid>FETCH-epo_espacenet_WO2024258068A13</originalsourceid><addsrcrecordid>eNrjZDBw9XF1Dgny9_N0VnAMCHAMcgwJDVZw9HNRcPb3A4r7KPi6hnj4uyiEeLgGubr5B_EwsKYl5hSn8kJpbgZlN9cQZw_d1IL8-NTigsTk1LzUkvhwfyMDIxMjUwsDMwtHQ2PiVAEA1hAnXg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>ELECTRONIC APPARATUS AND CONTROL METHOD THEREFOR</title><source>esp@cenet</source><creator>JEON, Yongkweon ; PARK, Kyungphil ; KIM, Hoyoung ; LEE, Chungman</creator><creatorcontrib>JEON, Yongkweon ; PARK, Kyungphil ; KIM, Hoyoung ; LEE, Chungman</creatorcontrib><description>Provided are an electronic apparatus and a control method therefor. The electronic apparatus comprises: a memory for storing at least one instruction; and at least one processor. The at least one processor executes the at least one instruction to: obtain a first quantization matrix expressed as a product of a first vector and a second vector for each of a plurality of layers on the basis of a weight for each of the plurality of layers included in a neural network model; obtain a second quantization matrix for each of the plurality of layers by multiplying each of first vectors of first quantization matrices corresponding to the plurality of layers by a second vector of a first quantization matrix corresponding to a previous layer; and quantize the neural network model by using the second quantization matrix. L'invention concerne un appareil électronique et son procédé de commande. L'appareil électronique comprend : une mémoire destinée à stocker au moins une instruction ; et au moins un processeur. Ledit au moins un processeur exécute ladite au moins une instruction pour : obtenir une première matrice de quantification exprimée en tant que produit d'un premier vecteur et d'un second vecteur pour chacune d'une pluralité de couches sur la base d'un poids pour chacune de la pluralité de couches comprises dans un modèle de réseau de neurones artificiels ; obtenir une seconde matrice de quantification pour chacune de la pluralité de couches par multiplication de chacun des premiers vecteurs de premières matrices de quantification correspondant à la pluralité de couches par un second vecteur d'une première matrice de quantification correspondant à une couche précédente ; et quantifier le modèle de réseau de neurones artificiels à l'aide de la seconde matrice de quantification. 전자 장치 및 이의 제어 방법이 제공된다. 본 전자 장치는 적어도 하나의 인스트럭션을 저장하는 메모리 및 적어도 하나의 프로세서를 포함한다. 적어도 하나의 프로세서는, 적어도 하나의 인스트럭션을 실행함으로써, 신경망 모델에 포함된 복수의 레이어 별 가중치에 기초하여 복수의 레이어 별로 제1 벡터 및 제2 벡터의 곱으로 표현된 제1 양자화 행렬을 획득하며, 복수의 레이어에 대응되는 제1 양자화 행렬들의 제1 벡터 각각을 이전 레이어에 대응되는 제1 양자화 행렬의 제2 벡터와 곱하여 복수의 레이어 별로 제2 양자화 행렬을 획득하며, 제2 양자화 행렬을 이용하여 신경망 모델을 양자화한다.</description><language>eng ; fre ; kor</language><subject>CALCULATING ; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS ; COMPUTING ; COUNTING ; PHYSICS</subject><creationdate>2024</creationdate><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&amp;date=20241219&amp;DB=EPODOC&amp;CC=WO&amp;NR=2024258068A1$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25542,76516</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&amp;date=20241219&amp;DB=EPODOC&amp;CC=WO&amp;NR=2024258068A1$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>JEON, Yongkweon</creatorcontrib><creatorcontrib>PARK, Kyungphil</creatorcontrib><creatorcontrib>KIM, Hoyoung</creatorcontrib><creatorcontrib>LEE, Chungman</creatorcontrib><title>ELECTRONIC APPARATUS AND CONTROL METHOD THEREFOR</title><description>Provided are an electronic apparatus and a control method therefor. The electronic apparatus comprises: a memory for storing at least one instruction; and at least one processor. The at least one processor executes the at least one instruction to: obtain a first quantization matrix expressed as a product of a first vector and a second vector for each of a plurality of layers on the basis of a weight for each of the plurality of layers included in a neural network model; obtain a second quantization matrix for each of the plurality of layers by multiplying each of first vectors of first quantization matrices corresponding to the plurality of layers by a second vector of a first quantization matrix corresponding to a previous layer; and quantize the neural network model by using the second quantization matrix. L'invention concerne un appareil électronique et son procédé de commande. L'appareil électronique comprend : une mémoire destinée à stocker au moins une instruction ; et au moins un processeur. Ledit au moins un processeur exécute ladite au moins une instruction pour : obtenir une première matrice de quantification exprimée en tant que produit d'un premier vecteur et d'un second vecteur pour chacune d'une pluralité de couches sur la base d'un poids pour chacune de la pluralité de couches comprises dans un modèle de réseau de neurones artificiels ; obtenir une seconde matrice de quantification pour chacune de la pluralité de couches par multiplication de chacun des premiers vecteurs de premières matrices de quantification correspondant à la pluralité de couches par un second vecteur d'une première matrice de quantification correspondant à une couche précédente ; et quantifier le modèle de réseau de neurones artificiels à l'aide de la seconde matrice de quantification. 전자 장치 및 이의 제어 방법이 제공된다. 본 전자 장치는 적어도 하나의 인스트럭션을 저장하는 메모리 및 적어도 하나의 프로세서를 포함한다. 적어도 하나의 프로세서는, 적어도 하나의 인스트럭션을 실행함으로써, 신경망 모델에 포함된 복수의 레이어 별 가중치에 기초하여 복수의 레이어 별로 제1 벡터 및 제2 벡터의 곱으로 표현된 제1 양자화 행렬을 획득하며, 복수의 레이어에 대응되는 제1 양자화 행렬들의 제1 벡터 각각을 이전 레이어에 대응되는 제1 양자화 행렬의 제2 벡터와 곱하여 복수의 레이어 별로 제2 양자화 행렬을 획득하며, 제2 양자화 행렬을 이용하여 신경망 모델을 양자화한다.</description><subject>CALCULATING</subject><subject>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2024</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZDBw9XF1Dgny9_N0VnAMCHAMcgwJDVZw9HNRcPb3A4r7KPi6hnj4uyiEeLgGubr5B_EwsKYl5hSn8kJpbgZlN9cQZw_d1IL8-NTigsTk1LzUkvhwfyMDIxMjUwsDMwtHQ2PiVAEA1hAnXg</recordid><startdate>20241219</startdate><enddate>20241219</enddate><creator>JEON, Yongkweon</creator><creator>PARK, Kyungphil</creator><creator>KIM, Hoyoung</creator><creator>LEE, Chungman</creator><scope>EVB</scope></search><sort><creationdate>20241219</creationdate><title>ELECTRONIC APPARATUS AND CONTROL METHOD THEREFOR</title><author>JEON, Yongkweon ; PARK, Kyungphil ; KIM, Hoyoung ; LEE, Chungman</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_WO2024258068A13</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng ; fre ; kor</language><creationdate>2024</creationdate><topic>CALCULATING</topic><topic>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>JEON, Yongkweon</creatorcontrib><creatorcontrib>PARK, Kyungphil</creatorcontrib><creatorcontrib>KIM, Hoyoung</creatorcontrib><creatorcontrib>LEE, Chungman</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>JEON, Yongkweon</au><au>PARK, Kyungphil</au><au>KIM, Hoyoung</au><au>LEE, Chungman</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>ELECTRONIC APPARATUS AND CONTROL METHOD THEREFOR</title><date>2024-12-19</date><risdate>2024</risdate><abstract>Provided are an electronic apparatus and a control method therefor. The electronic apparatus comprises: a memory for storing at least one instruction; and at least one processor. The at least one processor executes the at least one instruction to: obtain a first quantization matrix expressed as a product of a first vector and a second vector for each of a plurality of layers on the basis of a weight for each of the plurality of layers included in a neural network model; obtain a second quantization matrix for each of the plurality of layers by multiplying each of first vectors of first quantization matrices corresponding to the plurality of layers by a second vector of a first quantization matrix corresponding to a previous layer; and quantize the neural network model by using the second quantization matrix. L'invention concerne un appareil électronique et son procédé de commande. L'appareil électronique comprend : une mémoire destinée à stocker au moins une instruction ; et au moins un processeur. Ledit au moins un processeur exécute ladite au moins une instruction pour : obtenir une première matrice de quantification exprimée en tant que produit d'un premier vecteur et d'un second vecteur pour chacune d'une pluralité de couches sur la base d'un poids pour chacune de la pluralité de couches comprises dans un modèle de réseau de neurones artificiels ; obtenir une seconde matrice de quantification pour chacune de la pluralité de couches par multiplication de chacun des premiers vecteurs de premières matrices de quantification correspondant à la pluralité de couches par un second vecteur d'une première matrice de quantification correspondant à une couche précédente ; et quantifier le modèle de réseau de neurones artificiels à l'aide de la seconde matrice de quantification. 전자 장치 및 이의 제어 방법이 제공된다. 본 전자 장치는 적어도 하나의 인스트럭션을 저장하는 메모리 및 적어도 하나의 프로세서를 포함한다. 적어도 하나의 프로세서는, 적어도 하나의 인스트럭션을 실행함으로써, 신경망 모델에 포함된 복수의 레이어 별 가중치에 기초하여 복수의 레이어 별로 제1 벡터 및 제2 벡터의 곱으로 표현된 제1 양자화 행렬을 획득하며, 복수의 레이어에 대응되는 제1 양자화 행렬들의 제1 벡터 각각을 이전 레이어에 대응되는 제1 양자화 행렬의 제2 벡터와 곱하여 복수의 레이어 별로 제2 양자화 행렬을 획득하며, 제2 양자화 행렬을 이용하여 신경망 모델을 양자화한다.</abstract><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier
ispartof
issn
language eng ; fre ; kor
recordid cdi_epo_espacenet_WO2024258068A1
source esp@cenet
subjects CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
PHYSICS
title ELECTRONIC APPARATUS AND CONTROL METHOD THEREFOR
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-14T20%3A08%3A47IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-epo_EVB&rft_val_fmt=info:ofi/fmt:kev:mtx:patent&rft.genre=patent&rft.au=JEON,%20Yongkweon&rft.date=2024-12-19&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EWO2024258068A1%3C/epo_EVB%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true