서브메뉴
검색
Learning and Inference for Distributions: from Optimal Transport to Markov Chain Monte Carlo
Learning and Inference for Distributions: from Optimal Transport to Markov Chain Monte Carlo
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105516
- ISBN
- 9798263343286
- DDC
- 515
- 저자명
- Fan, Jiaojiao.
- 서명/저자
- Learning and Inference for Distributions: from Optimal Transport to Markov Chain Monte Carlo
- 발행사항
- [Sl] : Georgia Institute of Technology, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 134 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
- 주기사항
- Advisor: Chen, Yongxin.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2024.
- 초록/해제
- 요약Distributional data refers to data that provides insights into the probability distribution of a random variable or a set of random variables. In the machine learning (ML) domain, distributions are typically represented by a large number of samples. They have numer-ous crucial applications across various fields, including computer vision, healthcare and biomedical engineering. There is a famous quote: "Data is the food for machine learn-ing." Understanding how to transform and manipulate distributions in high-dimensional and data-driven formats is therefore essential. This dissertation focuses on large-scale dis-tributional problems, characterized either by the sheer volume of data or by high dimen-sionality.We investigate two fundamental mathematical tools: optimal transport and Markov Chain Monte Carlo (MCMC) sampling, both of which are integral to the transformation and manipulation of distributional data. Optimal transport is a centuries-old mathematical framework for comparing probability distributions. One of its most significant concepts, the Wasserstein distance, has had a substantial impact on machine learning, particularly in the study of Generative Adversarial Networks. Despite its important applications, such as calculating distribution discrepancies, interpolating, and aligning distributions, optimal transport is often hindered by its computational cost. In this thesis, we improve the com-putational efficiency of optimal transport from two perspectives. First, for discrete optimal transport (OT), where probability distributions are represented by probability vectors, we enhance the multi-marginal optimal transport (MOT) associated with graph structures. Sec-ond, we scale up OT to handle millions of samples by introducing neural OT solvers. Ad-ditionally, we demonstrate multiple downstream applications for these neural OT solvers, including Wasserstein gradient flow, Wasserstein barycenter and generalized geodesic.MCMC sampling remains the primary technique for sampling from a distribution and has numerous applications in Bayesian statistics, computational physics, and computationalbiology. However, it faces challenges in terms of computational scalability, particularly with increasing dimensions. To overcome these limitations, we introduce a novel algo-rithm based on the proximal sampler, and rigorously prove its computational complexity for converging to the target distribution. Additionally, we prove a Gaussian concentration inequality for semi-smooth functions, which could be of independent interest. This re-sult recovers the order of the well-known Gaussian concentration inequality for Lipschitz functions.
- 일반주제명
- Convex analysis
- 일반주제명
- Probability
- 일반주제명
- Visualization
- 일반주제명
- Entropy
- 일반주제명
- Neural networks
- 일반주제명
- Markov analysis
- 일반주제명
- Computer science
- 일반주제명
- Mathematics
- 기본자료저록
- Dissertations Abstracts International. 87-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2024 us c eng d■001000017360389
■00520260202105516
■006m o d
■007cr#unu||||||||
■020 ▼a9798263343286
■035 ▼a(MiAaPQ)AAI32309367
■035 ▼a(MiAaPQ)GeorgiaTech75659
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a515
■1001 ▼aFan, Jiaojiao.
■24510▼aLearning and Inference for Distributions: from Optimal Transport to Markov Chain Monte Carlo
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a134 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: B.
■500 ▼aAdvisor: Chen, Yongxin.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2024.
■520 ▼aDistributional data refers to data that provides insights into the probability distribution of a random variable or a set of random variables. In the machine learning (ML) domain, distributions are typically represented by a large number of samples. They have numer-ous crucial applications across various fields, including computer vision, healthcare and biomedical engineering. There is a famous quote: "Data is the food for machine learn-ing." Understanding how to transform and manipulate distributions in high-dimensional and data-driven formats is therefore essential. This dissertation focuses on large-scale dis-tributional problems, characterized either by the sheer volume of data or by high dimen-sionality.We investigate two fundamental mathematical tools: optimal transport and Markov Chain Monte Carlo (MCMC) sampling, both of which are integral to the transformation and manipulation of distributional data. Optimal transport is a centuries-old mathematical framework for comparing probability distributions. One of its most significant concepts, the Wasserstein distance, has had a substantial impact on machine learning, particularly in the study of Generative Adversarial Networks. Despite its important applications, such as calculating distribution discrepancies, interpolating, and aligning distributions, optimal transport is often hindered by its computational cost. In this thesis, we improve the com-putational efficiency of optimal transport from two perspectives. First, for discrete optimal transport (OT), where probability distributions are represented by probability vectors, we enhance the multi-marginal optimal transport (MOT) associated with graph structures. Sec-ond, we scale up OT to handle millions of samples by introducing neural OT solvers. Ad-ditionally, we demonstrate multiple downstream applications for these neural OT solvers, including Wasserstein gradient flow, Wasserstein barycenter and generalized geodesic.MCMC sampling remains the primary technique for sampling from a distribution and has numerous applications in Bayesian statistics, computational physics, and computationalbiology. However, it faces challenges in terms of computational scalability, particularly with increasing dimensions. To overcome these limitations, we introduce a novel algo-rithm based on the proximal sampler, and rigorously prove its computational complexity for converging to the target distribution. Additionally, we prove a Gaussian concentration inequality for semi-smooth functions, which could be of independent interest. This re-sult recovers the order of the well-known Gaussian concentration inequality for Lipschitz functions.
■590 ▼aSchool code: 0078.
■650 4▼aConvex analysis
■650 4▼aProbability
■650 4▼aVisualization
■650 4▼aEntropy
■650 4▼aNeural networks
■650 4▼aMarkov analysis
■650 4▼aComputer science
■650 4▼aMathematics
■690 ▼a0984
■690 ▼a0800
■690 ▼a0405
■690 ▼a0796
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05B.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360389▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


