본문

서브메뉴

Signal-to-Noise Ratio Aware Minimaxity and Its Asymptotic Expansion- [electronic resource]
Signal-to-Noise Ratio Aware Minimaxity and Its Asymptotic Expansion - [electronic resource...
Signal-to-Noise Ratio Aware Minimaxity and Its Asymptotic Expansion- [electronic resource]

상세정보

자료유형  
 학위논문파일 국외
최종처리일시  
20240214101920
ISBN  
9798380594721
DDC  
310
저자명  
Guo, Yilin.
서명/저자  
Signal-to-Noise Ratio Aware Minimaxity and Its Asymptotic Expansion - [electronic resource]
발행사항  
[S.l.]: : Columbia University., 2023
발행사항  
Ann Arbor : : ProQuest Dissertations & Theses,, 2023
형태사항  
1 online resource(192 p.)
주기사항  
Source: Dissertations Abstracts International, Volume: 85-04, Section: B.
주기사항  
Advisor: Maleki, Arian;Weng, Haolei.
학위논문주기  
Thesis (Ph.D.)--Columbia University, 2023.
사용제한주기  
This item must not be sold to any third party vendors.
초록/해제  
요약Since its development, the minimax framework has been one of the corner stones of theoretical statistics, and has contributed to the popularity of many well-known estimators, such as the regularized M-estimators for high-dimensional problems. In this thesis, we will first show through the example of sparse Gaussian sequence model, that the theoretical results under the classical minimax framework are insufficient for explaining empirical observations. In particular, both hard and soft thresholding estimators are (asymptotically) minimax, however, in practice they often exhibit sub-optimal performances at various signal-to-noise ratio (SNR) levels. To alleviate the discrepancy, we first demonstrate that this issue can be resolved if the signal-to-noise ratio is taken into account in the construction of the parameter space. We call the resulting minimax framework the signal-to-noise ratio aware minimaxity. Then, we showcase how one can use higher-order asymptotics to obtain accurate approximations of the SNR-aware minimax risk and discover minimax estimators. Theoretical findings obtained from this refined minimax framework provide new insights and practical guidance for the estimation of sparse signals.In a broader context, we investigated the same problem for sparse linear regression. We assume the random design and allow the feature matrix to be high dimensional as \uD835\uDC4B ∈ R\uD835\uDC5Bxp \uD835\uDC5D ≫ n. This adds an extra layer of challenge to the estimation of coefficients. Previous studies have largely relied on results expressed in rate-minimaxity, where estimators are compared based on minimax risk with order-wise accuracy, without specifying the precise constant in the approximation. This lack of precision contributes to the notable gap between theoretical conclusions of the asymptotic minimax estimators and empirical findings of the sub-optimality. This thesis addresses this gap by initially refining the classical minimax result, providing a characterization of the constant in the first-order approximation. Subsequently, by following the framework of SNR-aware minimaxity we introduced before, we derived improved approximations of minimax risks under different SNR levels. Notably, these refined results demonstrated better alignment with empirical findings compared to classical minimax outcomes. As showcased in the thesis, our enhanced SNR-aware minimax framework not only offers a more accurate depiction of sparse estimation but also unveils the crucial role of SNR in the problem. This insight emerges as a pivotal factor in assessing the optimality of estimators.
일반주제명  
Statistics.
일반주제명  
Electrical engineering.
키워드  
Linear regression
키워드  
Minimaxity
키워드  
Signal denoising
키워드  
Signal-to-noise ratio
키워드  
Sparsity
기타저자  
Columbia University Statistics
기본자료저록  
Dissertations Abstracts International. 85-04B.
기본자료저록  
Dissertation Abstract International
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008240612s2023      us  |||||||||||||||c||eng  d
■001000016935337
■00520240214101920
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798380594721
■035    ▼a(MiAaPQ)AAI30689933
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a310
■1001  ▼aGuo,  Yilin.
■24510▼aSignal-to-Noise  Ratio  Aware  Minimaxity  and  Its  Asymptotic  Expansion▼h[electronic  resource]
■260    ▼a[S.l.]:▼bColumbia  University.  ▼c2023
■260  1▼aAnn  Arbor  :▼bProQuest  Dissertations  &  Theses,  ▼c2023
■300    ▼a1  online  resource(192  p.)
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  85-04,  Section:  B.
■500    ▼aAdvisor:  Maleki,  Arian;Weng,  Haolei.
■5021  ▼aThesis  (Ph.D.)--Columbia  University,  2023.
■506    ▼aThis  item  must  not  be  sold  to  any  third  party  vendors.
■520    ▼aSince  its  development,  the  minimax  framework  has  been  one  of  the  corner  stones  of  theoretical  statistics,  and  has  contributed  to  the  popularity  of  many  well-known  estimators,  such  as  the  regularized  M-estimators  for  high-dimensional  problems.  In  this  thesis,  we  will  first  show  through  the  example  of  sparse  Gaussian  sequence  model,  that  the  theoretical  results  under  the  classical  minimax  framework  are  insufficient  for  explaining  empirical  observations.  In  particular,  both  hard  and  soft  thresholding  estimators  are  (asymptotically)  minimax,  however,  in  practice  they  often  exhibit  sub-optimal  performances  at  various  signal-to-noise  ratio  (SNR)  levels.  To  alleviate  the  discrepancy,  we  first  demonstrate  that  this  issue  can  be  resolved  if  the  signal-to-noise  ratio  is  taken  into  account  in  the  construction  of  the  parameter  space.  We  call  the  resulting  minimax  framework  the  signal-to-noise  ratio  aware  minimaxity.  Then,  we  showcase  how  one  can  use  higher-order  asymptotics  to  obtain  accurate  approximations  of  the  SNR-aware  minimax  risk  and  discover  minimax  estimators.  Theoretical  findings  obtained  from  this  refined  minimax  framework  provide  new  insights  and  practical  guidance  for  the  estimation  of  sparse  signals.In  a  broader  context,  we  investigated  the  same  problem  for  sparse  linear  regression.  We  assume  the  random  design  and  allow  the  feature  matrix  to  be  high  dimensional  as  \uD835\uDC4B  ∈  R\uD835\uDC5Bxp  \uD835\uDC5D  ≫  n.  This  adds  an  extra  layer  of  challenge  to  the  estimation  of  coefficients.  Previous  studies  have  largely  relied  on  results  expressed  in  rate-minimaxity,  where  estimators  are  compared  based  on  minimax  risk  with  order-wise  accuracy,  without  specifying  the  precise  constant  in  the  approximation.  This  lack  of  precision  contributes  to  the  notable  gap  between  theoretical  conclusions  of  the  asymptotic  minimax  estimators  and  empirical  findings  of  the  sub-optimality.  This  thesis  addresses  this  gap  by  initially  refining  the  classical  minimax  result,  providing  a  characterization  of  the  constant  in  the  first-order  approximation.  Subsequently,  by  following  the  framework  of  SNR-aware  minimaxity  we  introduced  before,  we  derived  improved  approximations  of  minimax  risks  under  different  SNR  levels.  Notably,  these  refined  results  demonstrated  better  alignment  with  empirical  findings  compared  to  classical  minimax  outcomes.  As  showcased  in  the  thesis,  our  enhanced  SNR-aware  minimax  framework  not  only  offers  a  more  accurate  depiction  of  sparse  estimation  but  also  unveils  the  crucial  role  of  SNR  in  the  problem.  This  insight  emerges  as  a  pivotal  factor  in  assessing  the  optimality  of  estimators.
■590    ▼aSchool  code:  0054.
■650  4▼aStatistics.
■650  4▼aElectrical  engineering.
■653    ▼aLinear  regression
■653    ▼aMinimaxity
■653    ▼aSignal  denoising
■653    ▼aSignal-to-noise  ratio
■653    ▼aSparsity
■690    ▼a0463
■690    ▼a0544
■71020▼aColumbia  University▼bStatistics.
■7730  ▼tDissertations  Abstracts  International▼g85-04B.
■773    ▼tDissertation  Abstract  International
■790    ▼a0054
■791    ▼aPh.D.
■792    ▼a2023
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T16935337▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.
■980    ▼a202402▼f2024

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
    TF06267 전자도서 마이폴더 부재도서신고 비도서대출신청

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

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.