서브메뉴
검색
상세정보
Generative Machine Learning, Granger Causality, and Optimal Intervention in Self-Exciting Spatiotemporal Processes.
Generative Machine Learning, Granger Causality, and Optimal Intervention in Self-Exciting Spatiotemporal Processes.
상세정보
- 자료유형
- 학위논문(국외)
- 기본표목-개인명
- 표제와 책임표시사항
- Generative Machine Learning, Granger Causality, and Optimal Intervention in Self-Exciting Spatiotemporal Processes.
- 발행, 배포, 간사 사항
- 발행, 배포, 간사 사항
- 형태사항
- 153 p.
- 일반주기
- Source: Dissertations Abstracts International, Volume: 87-03, Section: B.
- 일반주기
- Advisor: Banerjee, Moulinath;Sun, Yuekai.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 요약 등 주기
- 요약In this dissertation, we study self-exciting Hawkes processes where the occurrence of one event increases the likelihood of further events. In the first chapter, we propose a rigorous framework to study Granger causality in Spatial Hawkes networks and establish results about the recovery of causal links in the presence of noise in the observations. In the next part of the dissertation, we focus on problems at the interface of Hawkes processes and predictive policing. First, we propose a likelihood-free estimation method, leveraging Wasserstein GANs, for spatiotemporal Hawkes processes in the presence of missing events and demonstrate an application on crime hotspot forecasting in Bogota, Colombia. Next, we build optimization frameworks for understanding how strategic intervention at some nodes affects the overall dynamics of a spatial Hawkes network. Subsequently, we demonstrate the application of our proposed method for finding optimal patrolling sites across Los Angeles to reduce the spread of crimes under a low to moderate resource environment. Finally, the dissertation presents a shape-constrained nonparametric estimation method for Hawkes kernels under monotonicity or concavity constraints, along with applications to financial and seismological data. Together, these contributions make developments at the interface of point process modeling, generative AI, and optimization, aiding data-driven decision making in critical areas such as predictive policing and algorithmic threat detection.
- 주제명부출표목-일반주제명
- 비통제 색인어
- 비통제 색인어
- 비통제 색인어
- 비통제 색인어
- 비통제 색인어
- 부출표목-단체명
- 기본자료저록
- Dissertations Abstracts International. 87-03B.
- 전자적 위치 및 접속
- 원문정보보기
MARC
008260219s2025 us ||||||||||||||c||eng d■001000017359867
■00520260202105228
■006m o d
■007cr#unu||||||||
■020 ▼a9798291566879
■035 ▼a(MiAaPQ)AAI32271863
■035 ▼a(MiAaPQ)umichrackham006445
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a310
■1001 ▼aDas, Pramit.
■24510▼aGenerative Machine Learning, Granger Causality, and Optimal Intervention in Self-Exciting Spatiotemporal Processes.
■260 ▼a[S.l.]▼bUniversity of Michigan. ▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a153 p.
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: B.
■500 ▼aAdvisor: Banerjee, Moulinath;Sun, Yuekai.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aIn this dissertation, we study self-exciting Hawkes processes where the occurrence of one event increases the likelihood of further events. In the first chapter, we propose a rigorous framework to study Granger causality in Spatial Hawkes networks and establish results about the recovery of causal links in the presence of noise in the observations. In the next part of the dissertation, we focus on problems at the interface of Hawkes processes and predictive policing. First, we propose a likelihood-free estimation method, leveraging Wasserstein GANs, for spatiotemporal Hawkes processes in the presence of missing events and demonstrate an application on crime hotspot forecasting in Bogota, Colombia. Next, we build optimization frameworks for understanding how strategic intervention at some nodes affects the overall dynamics of a spatial Hawkes network. Subsequently, we demonstrate the application of our proposed method for finding optimal patrolling sites across Los Angeles to reduce the spread of crimes under a low to moderate resource environment. Finally, the dissertation presents a shape-constrained nonparametric estimation method for Hawkes kernels under monotonicity or concavity constraints, along with applications to financial and seismological data. Together, these contributions make developments at the interface of point process modeling, generative AI, and optimization, aiding data-driven decision making in critical areas such as predictive policing and algorithmic threat detection.
■590 ▼aSchool code: 0127.
■650 4▼aStatistics.
■653 ▼aPredictive policing
■653 ▼aNetwork optimization
■653 ▼aGenerative machine learning
■653 ▼aShape-restricted inference
■653 ▼aForecasting
■690 ▼a0463
■690 ▼a0796
■690 ▼a0800
■71020▼aUniversity of Michigan▼bStatistics.
■7730 ▼tDissertations Abstracts International▼g87-03B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359867▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
