서브메뉴
검색
Classical and Quantum Physics-Enhanced Machine Learning Algorithms in the Ordered and Chaotic Regimes- [electronic resource]
Classical and Quantum Physics-Enhanced Machine Learning Algorithms in the Ordered and Chaotic Regimes- [electronic resource]
상세정보
- 자료유형
- 학위논문
- Control Number
- 0016932880
- International Standard Book Number
- 9798379871154
- Dewey Decimal Classification Number
- 517
- Main Entry-Personal Name
- Holliday, Elliott Gregory.
- Publication, Distribution, etc. (Imprint
- [S.l.] : North Carolina State University., 2023
- Publication, Distribution, etc. (Imprint
- Ann Arbor : ProQuest Dissertations & Theses, 2023
- Physical Description
- 1 online resource(107 p.)
- General Note
- Source: Dissertations Abstracts International, Volume: 85-01, Section: B.
- General Note
- Advisor: Kumah, Divine;LeBlanc, Sharonda;Ruffino, Rico;Lindner, John F.;Daniels, Karen;Ditto, William L.
- Dissertation Note
- Thesis (Ph.D.)--North Carolina State University, 2023.
- Restrictions on Access Note
- This item must not be sold to any third party vendors.
- Summary, Etc.
- 요약Artificial neural networks (ANN) and machine learning have become critical for advancements in science, technology, and daily life. To expand the resources available to physicists for making discoveries and contributions to the field of physics, can we solve classical and quantum physics problems that exhibit both order and chaos using a neural network? Can we improve a neural network's ability to solve physics problems by giving it an internal physics intuition? Noting that calculus lies at the heart of both machine learning algorithms and physics, this thesis incorporates physics into the training process of ANNs to forecast classical Hamiltonian dynamical systems that exhibit both order and chaos such as the Henon-Heiles stellar potential, chaotic billiards, and the double pendulum. While the ANN is only given a singular formalism or set of constraints, what additional knowledge do we discover upon giving a physics formalism to the neural network? I find doing so recovers more about the system than what was inputted such as the energy, the dimensionality, and the fraction of chaotic orbits for a given energy range. While the Hamiltonian requires canonical coordinates, it also expands on the previous algorithm to forecast dynamics without canonical coordinates for the Lotka-Volterra predator-prey model and a video of a wooden pendulum clock. This thesis also develops this idea into quantum mechanics and explores the result of giving an ANN the Schrodinger equation so that it may recover eigenfunctions and energies. This method is tested on previously studied one- and twodimensional systems like the infinite square well and simple harmonic oscillator and two-dimensional infinite potential wells that classically exhibit order and chaos, such as elliptical, triangular, and cardioid-shaped wells. Physics-enhanced machine learning algorithms have the potential to improve how advances in physics and science are made but also could improve current ANNs by giving them scientific principles and knowledge.
- Subject Added Entry-Topical Term
- Calculus.
- Subject Added Entry-Topical Term
- Neurons.
- Subject Added Entry-Topical Term
- Physics.
- Subject Added Entry-Topical Term
- Partial differential equations.
- Subject Added Entry-Topical Term
- Neural networks.
- Subject Added Entry-Topical Term
- Eigenvalues.
- Subject Added Entry-Topical Term
- Energy.
- Subject Added Entry-Topical Term
- Billiards.
- Subject Added Entry-Topical Term
- Mathematics.
- Subject Added Entry-Topical Term
- Quantum physics.
- Added Entry-Corporate Name
- North Carolina State University.
- Host Item Entry
- Dissertations Abstracts International. 85-01B.
- Host Item Entry
- Dissertation Abstract International
- Electronic Location and Access
- 로그인을 한후 보실 수 있는 자료입니다.
- Control Number
- joongbu:639195
MARC
008240219s2023 ulk 00 kor■001000016932880
■00520240214100557
■006m o d
■007cr#unu||||||||
■020 ▼a9798379871154
■035 ▼a(MiAaPQ)AAI30516330
■035 ▼a(MiAaPQ)NCState_Univ18402040852
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a517
■1001 ▼aHolliday, Elliott Gregory.
■24510▼aClassical and Quantum Physics-Enhanced Machine Learning Algorithms in the Ordered and Chaotic Regimes▼h[electronic resource]
■260 ▼a[S.l.]▼bNorth Carolina State University. ▼c2023
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2023
■300 ▼a1 online resource(107 p.)
■500 ▼aSource: Dissertations Abstracts International, Volume: 85-01, Section: B.
■500 ▼aAdvisor: Kumah, Divine;LeBlanc, Sharonda;Ruffino, Rico;Lindner, John F.;Daniels, Karen;Ditto, William L.
■5021 ▼aThesis (Ph.D.)--North Carolina State University, 2023.
■506 ▼aThis item must not be sold to any third party vendors.
■520 ▼aArtificial neural networks (ANN) and machine learning have become critical for advancements in science, technology, and daily life. To expand the resources available to physicists for making discoveries and contributions to the field of physics, can we solve classical and quantum physics problems that exhibit both order and chaos using a neural network? Can we improve a neural network's ability to solve physics problems by giving it an internal physics intuition? Noting that calculus lies at the heart of both machine learning algorithms and physics, this thesis incorporates physics into the training process of ANNs to forecast classical Hamiltonian dynamical systems that exhibit both order and chaos such as the Henon-Heiles stellar potential, chaotic billiards, and the double pendulum. While the ANN is only given a singular formalism or set of constraints, what additional knowledge do we discover upon giving a physics formalism to the neural network? I find doing so recovers more about the system than what was inputted such as the energy, the dimensionality, and the fraction of chaotic orbits for a given energy range. While the Hamiltonian requires canonical coordinates, it also expands on the previous algorithm to forecast dynamics without canonical coordinates for the Lotka-Volterra predator-prey model and a video of a wooden pendulum clock. This thesis also develops this idea into quantum mechanics and explores the result of giving an ANN the Schrodinger equation so that it may recover eigenfunctions and energies. This method is tested on previously studied one- and twodimensional systems like the infinite square well and simple harmonic oscillator and two-dimensional infinite potential wells that classically exhibit order and chaos, such as elliptical, triangular, and cardioid-shaped wells. Physics-enhanced machine learning algorithms have the potential to improve how advances in physics and science are made but also could improve current ANNs by giving them scientific principles and knowledge.
■590 ▼aSchool code: 0155.
■650 4▼aCalculus.
■650 4▼aNeurons.
■650 4▼aPhysics.
■650 4▼aPartial differential equations.
■650 4▼aNeural networks.
■650 4▼aEigenvalues.
■650 4▼aEnergy.
■650 4▼aBilliards.
■650 4▼aMathematics.
■650 4▼aQuantum physics.
■690 ▼a0791
■690 ▼a0800
■690 ▼a0605
■690 ▼a0405
■690 ▼a0599
■71020▼aNorth Carolina State University.
■7730 ▼tDissertations Abstracts International▼g85-01B.
■773 ▼tDissertation Abstract International
■790 ▼a0155
■791 ▼aPh.D.
■792 ▼a2023
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T16932880▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
■980 ▼a202402▼f2024
