메뉴 건너뛰기
.. 내서재 .. 알림
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)
로그인 회원가입 고객센터 ENG
주제분류
인문학
인문학 일반 역사학 철학 종교학 어문학 기타 인문학
사회과학
사회과학 일반 경영학 관광학 교육학 군사학 문헌정보학 법학 사회학 신문방송학 심리과학 정치외교학 지리학 행정학 경제학 무역학 인류학 기타 사회과학
자연과학
자연과학 일반 물리학 생물학 생활과학 수학·통계학 천문학 지구과학 화학 기타 자연과학
공학
공학 일반 건축공학 기계공학 산업공학 재료·에너지공학 전기전자공학 조선해양공학 화학공학 컴퓨터학 기타공학
의약학
의학 일반 간호학 수의학 약학 한의학 기초의학 임상의학 기타 의약학
농수해양학
농수해양학 일반 농학 수산학 식품과학 임학 축산학 해상운송학 기타 농수해양학
예술체육학
예술 일반 디자인 미술 미용 사진 음악학 의상 체육 공연매체예술 기타 예술체육학
복합학
학제간연구 과학기술학 뇌인지과학 기술정책 기타 복합학

ENG

추천
검색

논문 기본 정보

자료유형
학위논문
저자정보

김태현 (연세대학교, 연세대학교 일반대학원)

지도교수
박유랑
발행연도
2023
저작권
연세대학교 논문은 저작권에 의해 보호받습니다.

이용수1

표지
AI에게 요청하기
추천
검색

이 논문의 연구 히스토리 (2)

2023

Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features
김태현 의생명시스템정보학과 2023.01 학위논문

2022

Personalized Progressive Federated Learning with Leveraging Client-Specific Vertical Features: Model Development and Validation
김태현 ,  장원석 ,  허순철 외 2명 한국지능정보시스템학회 학술대회논문집 2022.11 학술대회자료

초록· 키워드

오류제보하기
연합학습은 분산된 사용자들 사이에서 모델을 학습시키기 위한 방식이다. 하지만, 기존의 수평 연합 학습은 모델의 복잡성을 증가시키기 위해 수직 분할 데이터를 활용하지 못하며, 수직 연합 학습은 모든 사용자에게서 많은 양의 동일한 사용자가 공유 되어야한다. 반면, 연합 학습의 주요 과제 중 하나는 사용자 사이의 데이터 이질성과 독립-항등 분포가 아닌 환경에서의 학습이다. 본 연구에서 사용자 특징적인 수직 분할 데이터를 활용할 수 있는 다중 모델 기반 개인화된 알고리즘인 개인화된 점진 연합 학습 (Personalized Progressive Federated Learning, PPFL)을 제안한다. PPFL 의 성능은 Physionet Challenge 2012 와 eICU 및 세브란스 병원 데이터로 이루어의 현실 세계 데이터의 두 데이터에서 평가되었다. 병원 내 사망과 병원 체류 기간 예측의 두 가지 문제에 대해 정확도와 수신자 조작 특성 곡선 면적 (Area Under Receiver Operating Characteristic, AUROC)에 기반하여 평가하였다. PPFL 은 병원 내 사망 예측에서 평균 0.849 의 정확도와 0.790 의 AUROC 의 성능을 보여주었으며, 다른 비교 모델들에 비해 가장 높은 점수를 보여주었다. 체류 기간 예측에서 PPFL 은 평균 0.808 AUROC 로 비교 모델들 중 가장 높은 성능을 보였다.
#Personalized federated learning #vertically partitioned data #Non-IID data

목차

Contents
List of Figures ··········································································· iii
List of Tables ············································································· v
ABSTRACT ············································································· vi
Chapter 1. Introduction ·································································· 1
Chapter 2. Background ·································································· 4
2.1 Federated Learning and Design ·················································· 4
2.2 Federated Learning on Non-IID Data ··········································· 5
2.2 Federated with Medical Data····················································· 6
Chapter 3. Method ······································································· 7
3.1 Problem Formulation ······························································ 8
3.2 Horizontal Federated Learning ·················································· 9
3.3 Personalized Progressive Federated Learning ································ 10
3.3.1 Horizontal Network ························································· 10
3.3.2 Vertical Network ···························································· 12
3.3.3 Personalized Network······················································· 12
Chapter 4. Experiments································································· 17
4.1 Study Design ······································································ 17
4.2 Dataset ············································································· 18
4.3 Experiment Details ······························································· 20
4.4 Experiment Setting ······························································· 21
Chapter 5. Results ······································································· 22
5.1 Performance Evaluation ························································· 22
5.2 Effectiveness of Progressive Model ············································ 24
5.3 Performance Evaluation Using Real-World Clinical Data ·················· 26
Chapter 6. Discussion··································································· 28
Chapter 7. Conclusion ·································································· 32
References················································································ 33
Appendices ··············································································· 38
Abstract in Korean ······································································ 49

최근 본 자료

전체보기

댓글(0)

0