서브메뉴
검색
Large-Scale Optimization for Deep Neural Network Architecture: a Dynamical System Theory Perspective
Large-Scale Optimization for Deep Neural Network Architecture: a Dynamical System Theory Perspective
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105516
- ISBN
- 9798263341909
- DDC
- 519.7
- 저자명
- Liu, Guan-Horng.
- 서명/저자
- Large-Scale Optimization for Deep Neural Network Architecture: a Dynamical System Theory Perspective
- 발행사항
- [Sl] : Georgia Institute of Technology, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 173 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
- 주기사항
- Advisor: Theodorou, Evangelos A.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2024.
- 초록/해제
- 요약Optimization of deep neural networks (DNNs) has been a driving force in the advance-ment of modern machine learning and artificial intelligence. Despite efforts to design DNN architectures that leverage domain-specific knowledge, the development of optimization al-gorithms for training these million-parameter functions has often progressed independently of architectural innovations. This thesis delves into large-scale optimization methods that are not only aware of but also leverage the underlying deep architectural structures being optimized. Specifically, we demonstrate that the dynamical system and optimal control the-ory pave a profound foundation for algorithmic characterization in this unexplored avenue.Optimal control, in its broadest sense, examines the principle of optimization over dy-namical systems. This methodological perspective naturally arises in training neural dif-ferential equations and can be applied to standard DNNs by interpreting layer propaga-tion as discrete timesteps along a dynamical system, with Backpropagation emerging as an approximate dynamic programming method. Through development, we emphasize the significance of control-theoretic components such as differential programming, nonlinear Feynman-Kac, and path integral theory, which unify existing optimization methods while extending them to handle a broader class of complex dynamics and problem setups that can otherwise be hard to adapt or foresee.Our work demonstrates the broad applicability of control-theoretic optimization meth-ods in learning various deep architectures, including convolutional networks, neural ordi-nary differential equations, and neural stochastic differential equations such as denoising diffusion models. The resulting computational frameworks improve test-time performance and inference efficiency, enhance training robustness against unstable hyperparameters, ac-celerate convergence in terms of wall-clock time, and are applicable to a wide range of scientific problems, including image generation, restoration, translation, as well as solving mean-field games and opinion modeling.
- 일반주제명
- Dynamic programming
- 일반주제명
- Back propagation
- 일반주제명
- Decision making
- 일반주제명
- Neural networks
- 일반주제명
- Diffusion models
- 일반주제명
- Dynamical systems
- 일반주제명
- Visualization
- 일반주제명
- Games
- 일반주제명
- System theory
- 일반주제명
- Computer science
- 일반주제명
- Mathematics
- 기본자료저록
- Dissertations Abstracts International. 87-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2024 us c eng d■001000017360388
■00520260202105516
■006m o d
■007cr#unu||||||||
■020 ▼a9798263341909
■035 ▼a(MiAaPQ)AAI32309352
■035 ▼a(MiAaPQ)GeorgiaTech75655
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a519.7
■1001 ▼aLiu, Guan-Horng.
■24510▼aLarge-Scale Optimization for Deep Neural Network Architecture: a Dynamical System Theory Perspective
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a173 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: B.
■500 ▼aAdvisor: Theodorou, Evangelos A.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2024.
■520 ▼aOptimization of deep neural networks (DNNs) has been a driving force in the advance-ment of modern machine learning and artificial intelligence. Despite efforts to design DNN architectures that leverage domain-specific knowledge, the development of optimization al-gorithms for training these million-parameter functions has often progressed independently of architectural innovations. This thesis delves into large-scale optimization methods that are not only aware of but also leverage the underlying deep architectural structures being optimized. Specifically, we demonstrate that the dynamical system and optimal control the-ory pave a profound foundation for algorithmic characterization in this unexplored avenue.Optimal control, in its broadest sense, examines the principle of optimization over dy-namical systems. This methodological perspective naturally arises in training neural dif-ferential equations and can be applied to standard DNNs by interpreting layer propaga-tion as discrete timesteps along a dynamical system, with Backpropagation emerging as an approximate dynamic programming method. Through development, we emphasize the significance of control-theoretic components such as differential programming, nonlinear Feynman-Kac, and path integral theory, which unify existing optimization methods while extending them to handle a broader class of complex dynamics and problem setups that can otherwise be hard to adapt or foresee.Our work demonstrates the broad applicability of control-theoretic optimization meth-ods in learning various deep architectures, including convolutional networks, neural ordi-nary differential equations, and neural stochastic differential equations such as denoising diffusion models. The resulting computational frameworks improve test-time performance and inference efficiency, enhance training robustness against unstable hyperparameters, ac-celerate convergence in terms of wall-clock time, and are applicable to a wide range of scientific problems, including image generation, restoration, translation, as well as solving mean-field games and opinion modeling.
■590 ▼aSchool code: 0078.
■650 4▼aDynamic programming
■650 4▼aBack propagation
■650 4▼aDecision making
■650 4▼aNeural networks
■650 4▼aDiffusion models
■650 4▼aDynamical systems
■650 4▼aVisualization
■650 4▼aOrdinary differential equations
■650 4▼aGames
■650 4▼aSystem theory
■650 4▼aComputer science
■650 4▼aMathematics
■690 ▼a0800
■690 ▼a0984
■690 ▼a0405
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05B.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360388▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


