Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385892783> ?p ?o ?g. }
Showing items 1 to 66 of
66
with 100 items per page.
- W4385892783 endingPage "49" @default.
- W4385892783 startingPage "40" @default.
- W4385892783 abstract "최근 임베디드 환경에서 딥 러닝을 적용하고자 하는 요구가 증가하고 있다. 임베디드와 같은 제한적인 환경에서 딥 러닝 연산을 효율적으로 수행하기 위해서 Arm의 big.LITTLE과 같은 이기종 멀티코어 CPU 아키텍처가 널리 활용되고 있다. Arm은 딥 러닝 연산을 최적으로 수행하기 위해 Arm Compute Library(ACL)를 제공하고 있지만, big.LITTLE 구조를 가진 하드웨어의 잠재력을 충분히 활용하지는 못하고 있다. 본 논문은 각 하드웨어에 최적인 실행 커널과 스케줄을 자동으로 결정하기 위한 프로파일 기반 탐색 방법을 제안한다. 실험은 Tinker Edge R, Odroid N+, Snapdragon 865 HDK 보드에서 AlexNet, VGG16, MobileNetV2, GoogleNet 모델을 대상으로 진행하였으며, 모든 경우에서 제안된 방법이 기존의 방법보다 최대 266% 성능 향상을 보임을 확인하였다. 본 연구의 결과를 통해 임베디드 기기에서 저비용, 저전력, 고성능의 딥 러닝 수행이 가능할 것으로 기대한다." @default.
- W4385892783 created "2023-08-18" @default.
- W4385892783 creator A5023489409 @default.
- W4385892783 creator A5062763435 @default.
- W4385892783 creator A5080783573 @default.
- W4385892783 date "2023-07-31" @default.
- W4385892783 modified "2023-09-27" @default.
- W4385892783 title "Profile-based Optimization for Deep Learning on Heterogeneous Multi-core CPUs" @default.
- W4385892783 doi "https://doi.org/10.5573/ieie.2023.60.7.40" @default.
- W4385892783 hasPublicationYear "2023" @default.
- W4385892783 type Work @default.
- W4385892783 citedByCount "0" @default.
- W4385892783 crossrefType "journal-article" @default.
- W4385892783 hasAuthorship W4385892783A5023489409 @default.
- W4385892783 hasAuthorship W4385892783A5062763435 @default.
- W4385892783 hasAuthorship W4385892783A5080783573 @default.
- W4385892783 hasConcept C111919701 @default.
- W4385892783 hasConcept C118524514 @default.
- W4385892783 hasConcept C149635348 @default.
- W4385892783 hasConcept C150552126 @default.
- W4385892783 hasConcept C154945302 @default.
- W4385892783 hasConcept C162307627 @default.
- W4385892783 hasConcept C166957645 @default.
- W4385892783 hasConcept C173608175 @default.
- W4385892783 hasConcept C205649164 @default.
- W4385892783 hasConcept C2164484 @default.
- W4385892783 hasConcept C26771161 @default.
- W4385892783 hasConcept C2778828181 @default.
- W4385892783 hasConcept C41008148 @default.
- W4385892783 hasConcept C76155785 @default.
- W4385892783 hasConcept C78766204 @default.
- W4385892783 hasConceptScore W4385892783C111919701 @default.
- W4385892783 hasConceptScore W4385892783C118524514 @default.
- W4385892783 hasConceptScore W4385892783C149635348 @default.
- W4385892783 hasConceptScore W4385892783C150552126 @default.
- W4385892783 hasConceptScore W4385892783C154945302 @default.
- W4385892783 hasConceptScore W4385892783C162307627 @default.
- W4385892783 hasConceptScore W4385892783C166957645 @default.
- W4385892783 hasConceptScore W4385892783C173608175 @default.
- W4385892783 hasConceptScore W4385892783C205649164 @default.
- W4385892783 hasConceptScore W4385892783C2164484 @default.
- W4385892783 hasConceptScore W4385892783C26771161 @default.
- W4385892783 hasConceptScore W4385892783C2778828181 @default.
- W4385892783 hasConceptScore W4385892783C41008148 @default.
- W4385892783 hasConceptScore W4385892783C76155785 @default.
- W4385892783 hasConceptScore W4385892783C78766204 @default.
- W4385892783 hasIssue "7" @default.
- W4385892783 hasLocation W43858927831 @default.
- W4385892783 hasOpenAccess W4385892783 @default.
- W4385892783 hasPrimaryLocation W43858927831 @default.
- W4385892783 hasRelatedWork W1585350690 @default.
- W4385892783 hasRelatedWork W2000051442 @default.
- W4385892783 hasRelatedWork W2008876287 @default.
- W4385892783 hasRelatedWork W2074226157 @default.
- W4385892783 hasRelatedWork W2097819797 @default.
- W4385892783 hasRelatedWork W2357576365 @default.
- W4385892783 hasRelatedWork W2388757309 @default.
- W4385892783 hasRelatedWork W2488897859 @default.
- W4385892783 hasRelatedWork W2502560717 @default.
- W4385892783 hasRelatedWork W4245302940 @default.
- W4385892783 hasVolume "60" @default.
- W4385892783 isParatext "false" @default.
- W4385892783 isRetracted "false" @default.
- W4385892783 workType "article" @default.