서브메뉴
검색
Matrix Sketching in Optimization
Matrix Sketching in Optimization
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211152047
- ISBN
- 9798342103176
- DDC
- 330
- 저자명
- Dexter, Gregory.
- 서명/저자
- Matrix Sketching in Optimization
- 발행사항
- [Sl] : Purdue University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 199 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-04, Section: B.
- 주기사항
- Advisor: Drineas, Petros;Khanna, Rajiv.
- 학위논문주기
- Thesis (Ph.D.)--Purdue University, 2024.
- 초록/해제
- 요약Continuous optimization is a fundamental topic both in theoretical computer science and applications of machine learning. Meanwhile, an important idea in the development modern algorithms it the use of randomness to achieve empirical speedup and improved theoretical runtimes. Stochastic gradient descent (SGD) and matrix-multiplication time linear program solvers [1] are two important examples of such achievements. Matrix sketching and related ideas provide a theoretical framework for the behavior of random matrices and vectors that arise in these algorithms, thereby provide a natural way to better understand the behavior of such randomized algorithms. In this dissertation, we consider three general problems in this area:1. Linear Programming: Interior point methods (IPMs) are a common approach for solving linear programs (LPs) with strong theoretical guarantees and solid empirical performance. The time complexity of these methods is dominated by the cost of solv- ing a linear system of equations at each iteration. In common applications of linear programming, particularly in machine learning and scientific computing, the size of this linear system can become prohibitively large, requiring the use of iterative solvers, which provide an approximate solution to the linear system. However, approximately solving the linear system at each iteration of an IPM invalidates the theoretical guar- antees of common IPM analyses. To remedy this, we theoretically and empirically analyze (slightly modified) predictor-correctorIPMs when using approximate linear solvers: our approach guarantees that, when certain conditions are satisfied, the num- ber of IPM iterations does not increase and that the final solution remains feasible. We also provide practical instantiations of approximate linear solvers that satisfy these conditions for special classes of constraint matrices using randomized linear algebra.2. Eigenvalue Estimation: We study the problem of approximating the eigenspectrum of a symmetric matrix A ∈ R with bounded entries (i.e.. ||A||s≤ 1). We present a simple sublinear time algorithm that approximates all eigenvalues of A up to additive error ten using those of a randomly sampled()x() principal submatrix. Our result can be viewed as a concentration bound on the complete eigenspectrum of a random submatrix, significantly extending known bounds on just the singular values (the magnitudes of the eigenvalues). We give improved error bounds of ennz(A) and Ar when the rows of A can be sampled with probabilities proportional to their sparsities or their squared l2 norms respectively. Here nnz(A) is the number of non-zero entries in A and AF is its Frobenius norm. Even for the strictly easier problems of approximating the singular values or testing the existence of large negative eigenvalues [2], our results are the first that take advantage of non-uniform sampling to give improved error bounds. From a technical perspective, our results require several new eigenvalue concentration and perturbation bounds for matrices with bounded entries. Our non-uniform sampling bounds require a new algorithmic approach, which judiciously zeroes out entries of a randomly sampled submatrix to reduce variance, before computing the eigenvalues of that submatrix as estimates for those of A. We complement our theoretical results with numerical simulations, which demonstrate the effectiveness of our algorithms in practice.
- 일반주제명
- Sparsity
- 일반주제명
- Computer science
- 일반주제명
- Neural networks
- 일반주제명
- Probability
- 일반주제명
- Linear programming
- 일반주제명
- Data science
- 일반주제명
- Eigenvalues
- 일반주제명
- Optimization algorithms
- 일반주제명
- Linear algebra
- 기타저자
- Purdue University.
- 기본자료저록
- Dissertations Abstracts International. 86-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017162735
■00520250211152047
■006m o d
■007cr#unu||||||||
■020 ▼a9798342103176
■035 ▼a(MiAaPQ)AAI31345209
■035 ▼a(MiAaPQ)Purdue25655379
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a330
■1001 ▼aDexter, Gregory.
■24510▼aMatrix Sketching in Optimization
■260 ▼a[Sl]▼bPurdue University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a199 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-04, Section: B.
■500 ▼aAdvisor: Drineas, Petros;Khanna, Rajiv.
■5021 ▼aThesis (Ph.D.)--Purdue University, 2024.
■520 ▼aContinuous optimization is a fundamental topic both in theoretical computer science and applications of machine learning. Meanwhile, an important idea in the development modern algorithms it the use of randomness to achieve empirical speedup and improved theoretical runtimes. Stochastic gradient descent (SGD) and matrix-multiplication time linear program solvers [1] are two important examples of such achievements. Matrix sketching and related ideas provide a theoretical framework for the behavior of random matrices and vectors that arise in these algorithms, thereby provide a natural way to better understand the behavior of such randomized algorithms. In this dissertation, we consider three general problems in this area:1. Linear Programming: Interior point methods (IPMs) are a common approach for solving linear programs (LPs) with strong theoretical guarantees and solid empirical performance. The time complexity of these methods is dominated by the cost of solv- ing a linear system of equations at each iteration. In common applications of linear programming, particularly in machine learning and scientific computing, the size of this linear system can become prohibitively large, requiring the use of iterative solvers, which provide an approximate solution to the linear system. However, approximately solving the linear system at each iteration of an IPM invalidates the theoretical guar- antees of common IPM analyses. To remedy this, we theoretically and empirically analyze (slightly modified) predictor-correctorIPMs when using approximate linear solvers: our approach guarantees that, when certain conditions are satisfied, the num- ber of IPM iterations does not increase and that the final solution remains feasible. We also provide practical instantiations of approximate linear solvers that satisfy these conditions for special classes of constraint matrices using randomized linear algebra.2. Eigenvalue Estimation: We study the problem of approximating the eigenspectrum of a symmetric matrix A ∈ R with bounded entries (i.e.. ||A||s≤ 1). We present a simple sublinear time algorithm that approximates all eigenvalues of A up to additive error ten using those of a randomly sampled()x() principal submatrix. Our result can be viewed as a concentration bound on the complete eigenspectrum of a random submatrix, significantly extending known bounds on just the singular values (the magnitudes of the eigenvalues). We give improved error bounds of ennz(A) and Ar when the rows of A can be sampled with probabilities proportional to their sparsities or their squared l2 norms respectively. Here nnz(A) is the number of non-zero entries in A and AF is its Frobenius norm. Even for the strictly easier problems of approximating the singular values or testing the existence of large negative eigenvalues [2], our results are the first that take advantage of non-uniform sampling to give improved error bounds. From a technical perspective, our results require several new eigenvalue concentration and perturbation bounds for matrices with bounded entries. Our non-uniform sampling bounds require a new algorithmic approach, which judiciously zeroes out entries of a randomly sampled submatrix to reduce variance, before computing the eigenvalues of that submatrix as estimates for those of A. We complement our theoretical results with numerical simulations, which demonstrate the effectiveness of our algorithms in practice.
■590 ▼aSchool code: 0183.
■650 4▼aSparsity
■650 4▼aComputer science
■650 4▼aNeural networks
■650 4▼aProbability
■650 4▼aLinear programming
■650 4▼aData science
■650 4▼aEigenvalues
■650 4▼aOptimization algorithms
■650 4▼aLinear algebra
■690 ▼a0984
■690 ▼a0800
■71020▼aPurdue University.
■7730 ▼tDissertations Abstracts International▼g86-04B.
■790 ▼a0183
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17162735▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


