본문

서브메뉴

Computation and Estimation for Neural Networks via Log-Concave Coupling
Computation and Estimation for Neural Networks via Log-Concave Coupling
Computation and Estimation for Neural Networks via Log-Concave Coupling

상세정보

자료유형  
 학위논문 서양
최종처리일시  
20260202103031
ISBN  
9798286445271
DDC  
310
저자명  
McDonald, Curtis James.
서명/저자  
Computation and Estimation for Neural Networks via Log-Concave Coupling
발행사항  
[Sl] : Yale University, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
192 p
주기사항  
Source: Dissertations Abstracts International, Volume: 86-12, Section: B.
주기사항  
Advisor: Barron, Andrew R.
학위논문주기  
Thesis (Ph.D.)--Yale University, 2025.
초록/해제  
요약In this work, we consider a Bayesian method to train single-hidden-layer neural networks with ℓ1 controlled weights by defining posterior distributions using different subsets of the training data, and combining posterior means to form our estimators. We consider both a joint Bayesian model for all parameters of the neural network at once, and a greedy Bayes model training the neurons one at a time based on the residuals of previous fits.The log-likelihoods of the posterior distributions we define are multimodal and nonconcave, so sampling algorithms such as Markov Chain Monte Carlo (MCMC) will not be rapidly mixing to directly sample the posteriors. Using an auxiliary random variable, we produce a mixture distribution which we call a log-concave coupling. Using a continuous uniform prior over the ℓ1 ball, the conditional distributions of this mixture are log-concave, and the mixing distribution itself is log-concave when the number of parameters in our neural network exceeds the squared number of data points. Thus the mixture distribution can be sampled efficiently to produce samples for our original target density. For a discrete uniform prior over the ℓ1 ball intersected with a grid of small spacing, we study the performance of our posterior mean estimator in an arbitrary regret sense and a statistical risk sense. Say we have a target function g, with g˜ being its projection into the closure of the convex hull of signed neurons scaled by a constant. With neuron weight vectors of dimension d and N data points, we show an estimator defined by a combination of our posterior means in the joint sampling problem has arbitrary sequence regret and statistical risk within O([(log d)/N]1/4) of the regret and risk of g˜. For the greedy construction, the additional regret and risk is an improved third root power.
일반주제명  
Statistics
일반주제명  
Mathematics
일반주제명  
Theoretical mathematics
일반주제명  
Applied mathematics
키워드  
Bayesian methods
키워드  
Machine learning
키워드  
Markov Chain Monte Carlo
키워드  
Neural networks
키워드  
Statistical learning theory
기타저자  
Yale University Statistics and Data Science
기본자료저록  
Dissertations Abstracts International. 86-12B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017356768
■00520260202103031
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798286445271
■035    ▼a(MiAaPQ)AAI31845940
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a310
■1001  ▼aMcDonald,  Curtis  James.
■24510▼aComputation  and  Estimation  for  Neural  Networks  via  Log-Concave  Coupling
■260    ▼a[Sl]▼bYale  University▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a192  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  86-12,  Section:  B.
■500    ▼aAdvisor:  Barron,  Andrew  R.
■5021  ▼aThesis  (Ph.D.)--Yale  University,  2025.
■520    ▼aIn  this  work,  we  consider  a  Bayesian  method  to  train  single-hidden-layer  neural  networks  with  ℓ1  controlled  weights  by  defining  posterior  distributions  using  different  subsets  of  the  training  data,  and  combining  posterior  means  to  form  our  estimators.  We  consider  both  a  joint  Bayesian  model  for  all  parameters  of  the  neural  network  at  once,  and  a  greedy  Bayes  model  training  the  neurons  one  at  a  time  based  on  the  residuals  of  previous  fits.The  log-likelihoods  of  the  posterior  distributions  we  define  are  multimodal  and  nonconcave,  so  sampling  algorithms  such  as  Markov  Chain  Monte  Carlo  (MCMC)  will  not  be  rapidly  mixing  to  directly  sample  the  posteriors.  Using  an  auxiliary  random  variable,  we  produce  a  mixture  distribution  which  we  call  a  log-concave  coupling.  Using  a  continuous  uniform  prior  over  the  ℓ1  ball,  the  conditional  distributions  of  this  mixture  are  log-concave,  and  the  mixing  distribution  itself  is  log-concave  when  the  number  of  parameters  in  our  neural  network  exceeds  the  squared  number  of  data  points.  Thus  the  mixture  distribution  can  be  sampled  efficiently  to  produce  samples  for  our  original  target  density. For  a  discrete  uniform  prior  over  the  ℓ1  ball  intersected  with  a  grid  of  small  spacing,  we  study  the  performance  of  our  posterior  mean  estimator  in  an  arbitrary  regret  sense  and  a  statistical  risk  sense.  Say  we  have  a  target  function  g,  with  g˜  being  its  projection  into  the  closure  of  the  convex  hull  of  signed  neurons  scaled  by  a  constant.  With  neuron  weight  vectors  of  dimension  d  and  N  data  points,  we  show  an  estimator  defined  by  a  combination  of  our  posterior  means  in  the  joint  sampling  problem  has  arbitrary  sequence  regret  and  statistical  risk  within  O([(log  d)/N]1/4)  of  the  regret  and  risk  of  g˜.  For  the  greedy  construction,  the  additional  regret  and  risk  is  an  improved  third  root  power.
■590    ▼aSchool  code:  0265.
■650  4▼aStatistics
■650  4▼aMathematics
■650  4▼aTheoretical  mathematics
■650  4▼aApplied  mathematics
■653    ▼aBayesian  methods
■653    ▼aMachine  learning
■653    ▼aMarkov  Chain  Monte  Carlo
■653    ▼aNeural  networks
■653    ▼aStatistical  learning  theory
■690    ▼a0463
■690    ▼a0405
■690    ▼a0642
■690    ▼a0364
■71020▼aYale  University▼bStatistics  and  Data  Science.
■7730  ▼tDissertations  Abstracts  International▼g86-12B.
■790    ▼a0265
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17356768▼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
    TF16558 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.