서브메뉴
검색
Effective and Efficient Methods for Constrained Stochastic and Derivative-Free Optimization
Effective and Efficient Methods for Constrained Stochastic and Derivative-Free Optimization
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103639
- ISBN
- 9798314873328
- DDC
- 658
- 저자명
- Shi, Jiahao.
- 서명/저자
- Effective and Efficient Methods for Constrained Stochastic and Derivative-Free Optimization
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 226 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-11, Section: B.
- 주기사항
- Advisor: Berahas, Albert S.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약In this thesis, we develop, analyze, and implement effective and efficient methods for constrained stochastic and derivative-free optimization. For each method, we prove the convergence result under general assumptions. We also conduct numerical experiments to demonstrate the efficiency and efficacy of our proposed methods.In Chapter 2, we develop a model-based derivative-free optimization method. A key component of variational quantum algorithms (VQAs) is the choice of a classical optimizer to update the ansatz parameterization. Quantum algorithms, however, are destined to run on noisy hardware with limited fidelity, leading to objective function evaluations-such as those in the quantum approximate optimization algorithm (QAOA) or the variational quantum eigensolver (VQE)-that are compromised by both stochastic errors and hardware-induced noise. To address these challenges, we adapt recent developments from the noise-aware numerical optimization literature to these commonly used derivative-free model-based methods. We introduce the key defining characteristics of these novel noise-aware derivative-free model-based methods that separate them from standard model-based methods. We study an implementation of such noise-aware derivative-free model-based methods and compare its performance on demonstrative VQA simulations to classical solvers packaged in scikit-quant. Moreover, we propose a series of sequential quadratic programming (SQP) methods tailored for stochastic equality constrained optimization. In Chapter 3, we develop a stochastic method for solving equality constrained problems that utilizes predictive variance reduction within the SQP paradigm, and prove that a measure of first-order stationarity converges to zero in expectation under reasonable assumptions, with practical performance demonstrated on constrained binary classification problems. In Chapter 4, we propose a line search SQP method for general nonlinear equality constrained optimization that employs a modified line search strategy using second-order information to mitigate the Maratos effect, thereby achieving global convergence and local superlinear convergence. We further extend this approach to settings where the objective function is stochastic or represented as a finitesum, and design a practical inexact matrix-free variant and demonstrate the numerical performance empirically. Finally, in Chapter 5, we propose, analyze, and implement an SQP algorithm for minimizing a nonlinear smooth function subject to smooth equality constraints with noisy function and derivative estimates, addressing the challenges posed by bounded noise in the objective and constraint functions as well as potential rank-deficiency in the constraint Jacobians. This algorithm employs a decomposition approach to compute inexact search directions and utilizes adaptive step size selection schemes, which guarantee globalconvergence of infeasible stationarity error to a neighborhood of zero that is proportional to the noise level.
- 일반주제명
- Industrial engineering
- 일반주제명
- Applied mathematics
- 일반주제명
- Computer science
- 키워드
- Machine learning
- 기타저자
- University of Michigan Industrial & Operations Engineering
- 기본자료저록
- Dissertations Abstracts International. 86-11B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017358067
■00520260202103639
■006m o d
■007cr#unu||||||||
■020 ▼a9798314873328
■035 ▼a(MiAaPQ)AAI32092496
■035 ▼a(MiAaPQ)umichrackham006071
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a658
■1001 ▼aShi, Jiahao.
■24510▼aEffective and Efficient Methods for Constrained Stochastic and Derivative-Free Optimization
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a226 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-11, Section: B.
■500 ▼aAdvisor: Berahas, Albert S.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aIn this thesis, we develop, analyze, and implement effective and efficient methods for constrained stochastic and derivative-free optimization. For each method, we prove the convergence result under general assumptions. We also conduct numerical experiments to demonstrate the efficiency and efficacy of our proposed methods.In Chapter 2, we develop a model-based derivative-free optimization method. A key component of variational quantum algorithms (VQAs) is the choice of a classical optimizer to update the ansatz parameterization. Quantum algorithms, however, are destined to run on noisy hardware with limited fidelity, leading to objective function evaluations-such as those in the quantum approximate optimization algorithm (QAOA) or the variational quantum eigensolver (VQE)-that are compromised by both stochastic errors and hardware-induced noise. To address these challenges, we adapt recent developments from the noise-aware numerical optimization literature to these commonly used derivative-free model-based methods. We introduce the key defining characteristics of these novel noise-aware derivative-free model-based methods that separate them from standard model-based methods. We study an implementation of such noise-aware derivative-free model-based methods and compare its performance on demonstrative VQA simulations to classical solvers packaged in scikit-quant. Moreover, we propose a series of sequential quadratic programming (SQP) methods tailored for stochastic equality constrained optimization. In Chapter 3, we develop a stochastic method for solving equality constrained problems that utilizes predictive variance reduction within the SQP paradigm, and prove that a measure of first-order stationarity converges to zero in expectation under reasonable assumptions, with practical performance demonstrated on constrained binary classification problems. In Chapter 4, we propose a line search SQP method for general nonlinear equality constrained optimization that employs a modified line search strategy using second-order information to mitigate the Maratos effect, thereby achieving global convergence and local superlinear convergence. We further extend this approach to settings where the objective function is stochastic or represented as a finitesum, and design a practical inexact matrix-free variant and demonstrate the numerical performance empirically. Finally, in Chapter 5, we propose, analyze, and implement an SQP algorithm for minimizing a nonlinear smooth function subject to smooth equality constraints with noisy function and derivative estimates, addressing the challenges posed by bounded noise in the objective and constraint functions as well as potential rank-deficiency in the constraint Jacobians. This algorithm employs a decomposition approach to compute inexact search directions and utilizes adaptive step size selection schemes, which guarantee globalconvergence of infeasible stationarity error to a neighborhood of zero that is proportional to the noise level.
■590 ▼aSchool code: 0127.
■650 4▼aIndustrial engineering
■650 4▼aApplied mathematics
■650 4▼aComputer science
■653 ▼aNonlinear optimization
■653 ▼aStochastic optimization
■653 ▼aConstrained optimization
■653 ▼aDerivative-free optimization
■653 ▼aMachine learning
■690 ▼a0546
■690 ▼a0796
■690 ▼a0364
■690 ▼a0984
■71020▼aUniversity of Michigan▼bIndustrial & Operations Engineering.
■7730 ▼tDissertations Abstracts International▼g86-11B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358067▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


