서브메뉴
검색
Generalizations of the Classification Task
Generalizations of the Classification Task
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211150931
- ISBN
- 9798381977523
- DDC
- 658
- 저자명
- Cao, Alexander.
- 서명/저자
- Generalizations of the Classification Task
- 발행사항
- [Sl] : Northwestern University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 133 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 85-10, Section: A.
- 주기사항
- Advisor: Klabjan, Diego.
- 학위논문주기
- Thesis (Ph.D.)--Northwestern University, 2024.
- 초록/해제
- 요약Many machine learning problems are formulated into classification tasks. A typical classification framework, however, has strong, underlying assumptions that may not hold in real-world application. The goal is to relax these assumptions so as to generalize classification tasks to more nuanced settings. In this dissertation, I address two such assumptions by developing novel classifier learning methodologies. First, I study open-set recognition in which classifiers are trained on known classes but tested on additional unknown classes; recognizing known classes and vice-versa is the generalized classification task. As a solution, I introduce an autoencoder approach that is optimized to separate known and unknown classes in a useful manner. Second, I study early sequence classification. Most sequence classification methods assume the entire sequence is available during inference but here I consider the generalized, dynamic scenario of receiving elements of a sequence over time and classifying as soon as possible, as accurately as possible. Deciding when one has received enough elements is generally learned by exploration but my new method learns this directly in a supervised manner from the classifications itself.
- 일반주제명
- Industrial engineering
- 기타저자
- Northwestern University Industrial Engineering and Management Sciences
- 기본자료저록
- Dissertations Abstracts International. 85-10A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017160197
■00520250211150931
■006m o d
■007cr#unu||||||||
■020 ▼a9798381977523
■035 ▼a(MiAaPQ)AAI30990315
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a658
■1001 ▼aCao, Alexander.
■24510▼aGeneralizations of the Classification Task
■260 ▼a[Sl]▼bNorthwestern University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a133 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 85-10, Section: A.
■500 ▼aAdvisor: Klabjan, Diego.
■5021 ▼aThesis (Ph.D.)--Northwestern University, 2024.
■520 ▼aMany machine learning problems are formulated into classification tasks. A typical classification framework, however, has strong, underlying assumptions that may not hold in real-world application. The goal is to relax these assumptions so as to generalize classification tasks to more nuanced settings. In this dissertation, I address two such assumptions by developing novel classifier learning methodologies. First, I study open-set recognition in which classifiers are trained on known classes but tested on additional unknown classes; recognizing known classes and vice-versa is the generalized classification task. As a solution, I introduce an autoencoder approach that is optimized to separate known and unknown classes in a useful manner. Second, I study early sequence classification. Most sequence classification methods assume the entire sequence is available during inference but here I consider the generalized, dynamic scenario of receiving elements of a sequence over time and classifying as soon as possible, as accurately as possible. Deciding when one has received enough elements is generally learned by exploration but my new method learns this directly in a supervised manner from the classifications itself.
■590 ▼aSchool code: 0163.
■650 4▼aIndustrial engineering
■653 ▼aMachine learning problems
■653 ▼aAutoencoder approach
■653 ▼aClassification tasks
■690 ▼a0800
■690 ▼a0454
■690 ▼a0546
■71020▼aNorthwestern University▼bIndustrial Engineering and Management Sciences.
■7730 ▼tDissertations Abstracts International▼g85-10A.
■790 ▼a0163
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17160197▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
