본문

서브메뉴

PDE Methods for Deep Learning Analysis and Optimization
PDE Methods for Deep Learning Analysis and Optimization
PDE Methods for Deep Learning Analysis and Optimization

상세정보

자료유형  
 학위논문 서양
최종처리일시  
20260202105602
ISBN  
9798263395636
DDC  
541.34
저자명  
Sun, Yuxin.
서명/저자  
PDE Methods for Deep Learning Analysis and Optimization
발행사항  
[Sl] : Georgia Institute of Technology, 2024
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2024
형태사항  
122 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
주기사항  
Advisor: Yezzi, Anthony J.;Sundaramoorthi, Ganesh.
학위논문주기  
Thesis (Ph.D.)--Georgia Institute of Technology, 2024.
초록/해제  
요약The objective of this dissertation is to use tools from partial differential equations (PDEs) to understand and construct deep learning algorithms. This research includes the theory-inspired design of optimization algorithms for deep learning, theoretical analysis of deep network training, and applications in computer vision. We introduce a recently developed framework (PDE acceleration), which is a variational approach to accelerated optimization with PDEs, in the context of optimization of deep networks, which leads to a novel and simple extension of stochastic gradient descent (SGD) with momentum. We empirically validate the theory and evaluate our new algorithm on image classification showing empirical improvement over SGD. To further enhance the performance of deep learning algorithms, we need a better understanding of the stability and convergence properties. We discovered restrained numerical instabilities in current training practices of deep networks. To explain this phenomenon, we present a theoretical framework using numerical analysis of PDE and analyzing the gradient descent PDE of a simplified convolutional neural network (CNN). We also link restrained instabilities to the recently discovered Edge of Stability (EoS) phenomena and provide new insights and predictions about the EoS. Further, the special potential of "geometric" PDEs in particular to advance deep learning applications is explored in this dissertation. Under the geometric PDE's framework, we provide a theoretical analysis to understand the instability caused by the Eikonal loss and explain how some existing approaches can unknowingly mitigate this instability. Furthermore, those regularization enables the use of new neural networks with higher representation power that can capture finer scale details of shape. In summary, we believe the tools we've introduced could improve deep learning practice.
일반주제명  
Diffusion
일반주제명  
Partial differential equations
일반주제명  
Deep learning
일반주제명  
Computer vision
일반주제명  
Ordinary differential equations
일반주제명  
Neural networks
일반주제명  
Computer science
일반주제명  
Mathematics
기타저자  
Georgia Institute of Technology.
기본자료저록  
Dissertations Abstracts International. 87-05B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2024        us                              c    eng  d
■001000017360658
■00520260202105602
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798263395636
■035    ▼a(MiAaPQ)AAI32315993
■035    ▼a(MiAaPQ)GeorgiaTech76924
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a541.34
■1001  ▼aSun,  Yuxin.
■24510▼aPDE  Methods  for  Deep  Learning  Analysis  and  Optimization
■260    ▼a[Sl]▼bGeorgia  Institute  of  Technology▼c2024
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2024
■300    ▼a122  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-05,  Section:  B.
■500    ▼aAdvisor:  Yezzi,  Anthony  J.;Sundaramoorthi,  Ganesh.
■5021  ▼aThesis  (Ph.D.)--Georgia  Institute  of  Technology,  2024.
■520    ▼aThe  objective  of  this  dissertation  is  to  use  tools  from  partial  differential  equations  (PDEs)  to  understand  and  construct  deep  learning  algorithms.  This  research  includes  the  theory-inspired  design  of  optimization  algorithms  for  deep  learning,  theoretical  analysis  of  deep  network  training,  and  applications  in  computer  vision.  We  introduce  a  recently  developed  framework  (PDE  acceleration),  which  is  a  variational  approach  to  accelerated  optimization  with  PDEs,  in  the  context  of  optimization  of  deep  networks,  which  leads  to  a  novel  and  simple  extension  of  stochastic  gradient  descent  (SGD)  with  momentum.  We  empirically  validate  the  theory  and  evaluate  our  new  algorithm  on  image  classification  showing  empirical  improvement  over  SGD.  To  further  enhance  the  performance  of  deep  learning  algorithms,  we  need  a  better  understanding  of  the  stability  and  convergence  properties.  We  discovered  restrained  numerical  instabilities  in  current  training  practices  of  deep  networks.  To  explain  this  phenomenon,  we  present  a  theoretical  framework  using  numerical  analysis  of  PDE  and  analyzing  the  gradient  descent  PDE  of  a  simplified  convolutional  neural  network  (CNN).  We  also  link  restrained  instabilities  to  the  recently  discovered  Edge  of  Stability  (EoS)  phenomena  and  provide  new  insights  and  predictions  about  the  EoS.  Further,  the  special  potential  of  "geometric"  PDEs  in  particular  to  advance  deep  learning  applications  is  explored  in  this  dissertation.  Under  the  geometric  PDE's  framework,  we  provide  a  theoretical  analysis  to  understand  the  instability  caused  by  the  Eikonal  loss  and  explain  how  some  existing  approaches  can  unknowingly  mitigate  this  instability.  Furthermore,  those  regularization  enables  the  use  of  new  neural  networks  with  higher  representation  power  that  can  capture  finer  scale  details  of  shape.  In  summary,  we  believe  the  tools  we've  introduced  could  improve  deep  learning  practice.
■590    ▼aSchool  code:  0078.
■650  4▼aDiffusion
■650  4▼aPartial  differential  equations
■650  4▼aDeep  learning
■650  4▼aComputer  vision
■650  4▼aOrdinary  differential  equations
■650  4▼aNeural  networks
■650  4▼aComputer  science
■650  4▼aMathematics
■690    ▼a0984
■690    ▼a0800
■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=T17360658▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

미리보기

내보내기

chatGPT토론

Ai 추천 관련 도서


    신착도서 더보기
    최근 3년간 통계입니다.

    소장정보

    • 예약
    • 소재불명신고
    • 나의폴더
    • 우선정리요청
    • 비도서대출신청
    • 야간 도서대출신청
    소장자료
    등록번호 청구기호 소장처 대출가능여부 대출정보
    TF17465 전자도서 대출가능 마이폴더 부재도서신고 비도서대출신청 야간 도서대출신청

    * 대출중인 자료에 한하여 예약이 가능합니다. 예약을 원하시면 예약버튼을 클릭하십시오.

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.