서브메뉴
검색
Robust and Efficient Deep Learning Under Data Constraints
Robust and Efficient Deep Learning Under Data Constraints
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260311091500.5
- ISBN
- 9798270234560
- DDC
- 005
- 저자명
- Chen, Xuxi
- 서명/저자
- Robust and Efficient Deep Learning Under Data Constraints / Xuxi Chen
- 발행사항
- [Sl] : The University of Texas at Austin, 2025
- 형태사항
- 1 electronic resource (234 pages)
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-06, Section: B.
- 주기사항
- Advisors: Wang, Zhangyang Atlas Committee members: Cheng, Yu; Ding, Ying; Kim, Hyeji; Zhu, Hao.
- 학위논문주기
- - Ph.D. : The University of Texas at Austin, 2025.
- 초록/해제
- 요약In an era where deep learning is increasingly integrated across a wide range of domains, from scientific research to everyday applications, the foundational role of data has become more critical than ever. However, the growing demand for high-quality data, coupled with its limited availability and the rising computational costs associated with large-scale datasets, presents substantial challenges to the scalable and sustainable deployment of advanced deep learning methodologies.The overarching goal of this thesis is to develop and advance machine learning techniques that remain robust and efficient under various data constraints. Specifically, it addresses seemingly opposite yet connected challenges: (1) the scarcity of high-quality data, and (2) the overwhelming volume of data in diverse learning scenarios. The first part of the thesis focuses on strategies to mitigate the impact of limited high-quality data in both conventional and scientific machine learning contexts. Particularly, we investigate the use of sparse models in conjunction with tailored optimization techniques to enhance performance in low-data regimes, demonstrating how these approaches can yield significant improvements in model accuracy and generalization. The second part shifts attention to the challenges posed by data abundance. Here, we introduce novel frameworks that employ techniques of data selection and compression to facilitate efficient training on large datasets, enabling models to achieve comparable performance with fewer data samples or less computational resources. Additionally, we examine a special class of problems, i.e., game theory, where the data volume can be indefinite. We propose a new framework that supports the efficient game solving, achieving faster convergence while maintaining solution quality.
- 언어주기
- English
- 일반주제명
- Materials science
- 키워드
- Deep learning
- 키워드
- Data selection
- 기타저자
- The University of Texas at Austin Electrical and Computer Engineering
- 기본자료저록
- Dissertations Abstracts International. 87-06B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260311s2025 us eng d■001000017361270
■00520260311091500.5
■006m o d
■007cr|nu||||||||
■020 ▼a9798270234560
■040 ▼aMiAaPQD▼beng▼cMiAaPQD▼erda
■082 ▼a005
■1001 ▼aChen, Xuxi▼eauthor.
■24510▼aRobust and Efficient Deep Learning Under Data Constraints ▼cXuxi Chen
■260 ▼a[Sl]▼bThe University of Texas at Austin▼c2025
■264 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a1 electronic resource (234 pages)
■336 ▼atext▼btxt▼2rdacontent
■337 ▼acomputer▼bc▼2rdamedia
■338 ▼aonline resource▼bcr▼2rdacarrier
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-06, Section: B.
■500 ▼aAdvisors: Wang, Zhangyang Atlas Committee members: Cheng, Yu; Ding, Ying; Kim, Hyeji; Zhu, Hao.
■5021 ▼bPh.D.▼cThe University of Texas at Austin▼d2025.
■520 ▼aIn an era where deep learning is increasingly integrated across a wide range of domains, from scientific research to everyday applications, the foundational role of data has become more critical than ever. However, the growing demand for high-quality data, coupled with its limited availability and the rising computational costs associated with large-scale datasets, presents substantial challenges to the scalable and sustainable deployment of advanced deep learning methodologies.The overarching goal of this thesis is to develop and advance machine learning techniques that remain robust and efficient under various data constraints. Specifically, it addresses seemingly opposite yet connected challenges: (1) the scarcity of high-quality data, and (2) the overwhelming volume of data in diverse learning scenarios. The first part of the thesis focuses on strategies to mitigate the impact of limited high-quality data in both conventional and scientific machine learning contexts. Particularly, we investigate the use of sparse models in conjunction with tailored optimization techniques to enhance performance in low-data regimes, demonstrating how these approaches can yield significant improvements in model accuracy and generalization. The second part shifts attention to the challenges posed by data abundance. Here, we introduce novel frameworks that employ techniques of data selection and compression to facilitate efficient training on large datasets, enabling models to achieve comparable performance with fewer data samples or less computational resources. Additionally, we examine a special class of problems, i.e., game theory, where the data volume can be indefinite. We propose a new framework that supports the efficient game solving, achieving faster convergence while maintaining solution quality.
■546 ▼aEnglish
■590 ▼aSchool code: 0227
■650 4▼aMaterials science
■653 ▼aDeep learning
■653 ▼aMachine learning techniques
■653 ▼aData selection
■7102 ▼aThe University of Texas at Austin▼bElectrical and Computer Engineering.▼edegree granting institution.
■7201 ▼aWang, Zhangyang Atlas▼edegree supervisor.
■7730 ▼tDissertations Abstracts International▼g87-06B.
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17361270▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
