서브메뉴
검색
Theory and Computation of Wasserstein Geometric Flows with Application to Time-Dependent Schrodinger Equation
Theory and Computation of Wasserstein Geometric Flows with Application to Time-Dependent Schrodinger Equation
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260209102913
- ISBN
- 9798265406620
- DDC
- 530
- 저자명
- Wu, Hao.
- 서명/저자
- Theory and Computation of Wasserstein Geometric Flows with Application to Time-Dependent Schrodinger Equation
- 발행사항
- [Sl] : Georgia Institute of Technology, 2023
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2023
- 형태사항
- 112 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
- 주기사항
- Advisor: Zhou, Haomin.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2023.
- 초록/해제
- 요약This dissertation focuses on the systematic study of a novel computational framework to solve Wasserstein geometric flows. In particular, two important classes of such flows, Wasserstein gradient flows (WGFs) and Wasserstein Hamiltonian flows (WHFs), are the main examples used throughout this research-they have many applications in real-world physics systems and more recently in deep learning problems such as generative models. Wasserstein geometric flows describe time evolution of probability density functions on the infinite-dimensional Wasserstein manifold. Therefore, it is critical to develop fast, accurate, and scalable numerical schemes to solves these flows.The main feature of our computational framework is to use general reduced-order models, such as deep neural networks, to parameterize the push-forward maps such that they can push a simple reference density to the ones solving WGFs or WHFs. This approach essentially reduces these flows defined on infinite-dimensional Wasserstein mainfold to finite-dimensional dynamical systems of the parameters of reduced-order models. These new dynamical systems are parameterizations of the WGFs and WHFs, which we call PWGFs and PWHFs for short, and are defined on the finite-dimensional parameter space. By leveraging a pullback Wasserstein metric on the parameter space, we can develop effective numerical methods to approximate the solutions of these flows.For WGFs, we show that our proposed PWGF scheme can be applied to general linear, nonlinear, interactive functionals or a combination of them defined on Wasserstein manifolds. Moreover, our scheme does not require any spatial discretization and thus is scalable to cases where the space dimensions of the problems are high. Our approach does not need to solve any nonconvex optimization problems but only require solving standard least squares problems in each time step. With these features, PWGF demonstrates promising computational efficiency and accuracy on a variety of WGF examples, as shown in our numerical experiments. A comprehensive analysis of the approximation errors using PWGFis also provided in this work, which lays the foundation theoretically.For WHFs, we adopt the similar idea but apply it to the more challenging Hamiltonian systems on Wasserstein manifolds. We propose a proper way to parameterize the adjoint variable in the parameter space. To preserve the Hamiltonian, we employ a sympletic numerical scheme to solve the PWHF, where a fixed-point iteration scheme is used to solve the implicit update equation of the model parameters. Similar to PWGF, PWHF does not require solving nonconvex optimization problems and thus avoids all issues of existing nonconvex optimization algorithms. We present the connection between the Lagrangian and Eulerian perspectives of the original flows using PWHF. Approximation error analysis and a number of numerical examples are provided using PWHF. Furthermore, we consider the Schrodinger equations (SEs) as operator evolution equations, and show how to use 짢 PWHF to solve them. Numerical results are also provided to demonstrate its promising performance in solving SEs where the state spaces are of high dimension.
- 일반주제명
- Schrodinger equation
- 일반주제명
- Error analysis
- 일반주제명
- Dynamical systems
- 일반주제명
- Neural networks
- 일반주제명
- Mathematics
- 일반주제명
- Theoretical physics
- 기본자료저록
- Dissertations Abstracts International. 87-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260203s2023 us c eng d■001000017366007
■00520260209102913
■006m o d
■007cr#unu||||||||
■020 ▼a9798265406620
■035 ▼a(MiAaPQ)AAI32316164
■035 ▼a(MiAaPQ)GeorgiaTech72711
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a530
■1001 ▼aWu, Hao.
■24510▼aTheory and Computation of Wasserstein Geometric Flows with Application to Time-Dependent Schrodinger Equation
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2023
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2023
■300 ▼a112 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: B.
■500 ▼aAdvisor: Zhou, Haomin.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2023.
■520 ▼aThis dissertation focuses on the systematic study of a novel computational framework to solve Wasserstein geometric flows. In particular, two important classes of such flows, Wasserstein gradient flows (WGFs) and Wasserstein Hamiltonian flows (WHFs), are the main examples used throughout this research-they have many applications in real-world physics systems and more recently in deep learning problems such as generative models. Wasserstein geometric flows describe time evolution of probability density functions on the infinite-dimensional Wasserstein manifold. Therefore, it is critical to develop fast, accurate, and scalable numerical schemes to solves these flows.The main feature of our computational framework is to use general reduced-order models, such as deep neural networks, to parameterize the push-forward maps such that they can push a simple reference density to the ones solving WGFs or WHFs. This approach essentially reduces these flows defined on infinite-dimensional Wasserstein mainfold to finite-dimensional dynamical systems of the parameters of reduced-order models. These new dynamical systems are parameterizations of the WGFs and WHFs, which we call PWGFs and PWHFs for short, and are defined on the finite-dimensional parameter space. By leveraging a pullback Wasserstein metric on the parameter space, we can develop effective numerical methods to approximate the solutions of these flows.For WGFs, we show that our proposed PWGF scheme can be applied to general linear, nonlinear, interactive functionals or a combination of them defined on Wasserstein manifolds. Moreover, our scheme does not require any spatial discretization and thus is scalable to cases where the space dimensions of the problems are high. Our approach does not need to solve any nonconvex optimization problems but only require solving standard least squares problems in each time step. With these features, PWGF demonstrates promising computational efficiency and accuracy on a variety of WGF examples, as shown in our numerical experiments. A comprehensive analysis of the approximation errors using PWGFis also provided in this work, which lays the foundation theoretically.For WHFs, we adopt the similar idea but apply it to the more challenging Hamiltonian systems on Wasserstein manifolds. We propose a proper way to parameterize the adjoint variable in the parameter space. To preserve the Hamiltonian, we employ a sympletic numerical scheme to solve the PWHF, where a fixed-point iteration scheme is used to solve the implicit update equation of the model parameters. Similar to PWGF, PWHF does not require solving nonconvex optimization problems and thus avoids all issues of existing nonconvex optimization algorithms. We present the connection between the Lagrangian and Eulerian perspectives of the original flows using PWHF. Approximation error analysis and a number of numerical examples are provided using PWHF. Furthermore, we consider the Schrodinger equations (SEs) as operator evolution equations, and show how to use 짢 PWHF to solve them. Numerical results are also provided to demonstrate its promising performance in solving SEs where the state spaces are of high dimension.
■590 ▼aSchool code: 0078.
■650 4▼aSchrodinger equation
■650 4▼aError analysis
■650 4▼aDynamical systems
■650 4▼aOrdinary differential equations
■650 4▼aNeural networks
■650 4▼aMathematics
■650 4▼aTheoretical physics
■690 ▼a0800
■690 ▼a0405
■690 ▼a0753
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05B.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2023
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17366007▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


