메뉴 건너뛰기
소속 기관 / 학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
고객센터 ENG
주제분류

논문 기본 정보

저자정보
(충북대학교) (국립경찰대학) (충북대학교)
저널정보
제어로봇시스템학회 제어로봇시스템학회 논문지 제어로봇시스템학회 논문지 제29권 제12호
오류 신고하기

피인용 0

검색

    초록·키워드

    Network pruning is a widely used technique for reducing the memory and computational requirements of deep neural networks while minimizing performance degradation. Structured pruning, in particular, offers an effective approach without the need for additional hardware. However, many structured pruning methods require pre-trained large networks, increasing the overall training time. To address this challenge, this study proposes structured pruning using the early pruning indicator (EPI) to enable network compression from the initial training phase. Despite its effectiveness, this technique is sensitive to hyperparameter selection when applied to small datasets. This study analyzes hyperparameters for the application of EPI-based structured pruning to small datasets such as CIFAR10 and CIFAR100. By identifying optimal hyperparameters, the proposed method achieves performance scores of 92.19% and 69.27% on the CIFAR10 and CIFAR100 datasets, respectively, with a compression rate of 70% using the RestNet56 network.

    최근 본 자료 전체보기