BROWSE

Related Scientist

Kim, Sundong's photo.

Kim, Sundong
데이터 사이언스 그룹
more info

ITEM VIEW & DOWNLOAD

Ada-boundary: accelerating DNN training via adaptive boundary batch selection

Cited 0 time in webofscience Cited 0 time in scopus
6 Viewed 0 Downloaded
Title
Ada-boundary: accelerating DNN training via adaptive boundary batch selection
Author(s)
Hwanjun Song; Sundong Kim; Minseok Kim; Jae‑Gil Lee
Publication Date
2020-09
Journal
MACHINE LEARNING, v.109, no.9-10, pp.1837 - 1853
Publisher
SPRINGER
Abstract
© The Author(s), under exclusive licence to Springer Science+Business Media LLC, part of Springer Nature. Neural networks converge faster with help from a smart batch selection strategy. In this regard, we proposeAda-Boundary, a novel and simple adaptive batch selection algorithm that constructs an effective mini-batch according to the learning progress of the model. Our key idea is to exploitconfusingsamples for which the model cannot predict labels with high confidence. Thus, samples near the current decision boundary are considered to be the most effective for expediting convergence. Taking advantage of this design,Ada-Boundarymaintained its dominance for various degrees of training difficulty. We demonstrate the advantage ofAda-Boundaryby extensive experimentation using CNNs with five benchmark data sets.Ada-Boundarywas shown to produce a relative improvement in test errors by up to 31.80% compared with the baseline for a fixed wall-clock training time, thereby achieving a faster convergence speed
URI
https://pr.ibs.re.kr/handle/8788114/7653
DOI
10.1007/s10994-020-05903-6
ISSN
0885-6125
Appears in Collections:
Pioneer Research Center for Mathematical and Computational Sciences(수리 및 계산과학 연구단) > Data Science Group(데이터 사이언스 그룹) > 1. Journal Papers (저널논문)
Files in This Item:
There are no files associated with this item.

qrcode

  • facebook

    twitter

  • Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
해당 아이템을 이메일로 공유하기 원하시면 인증을 거치시기 바랍니다.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse