서브메뉴
검색
Deep Probabilistic Modeling for Causal Inference and Decision Making
Deep Probabilistic Modeling for Causal Inference and Decision Making
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103118
- ISBN
- 9798288861840
- DDC
- 574
- 저자명
- Wu, Yulun.
- 서명/저자
- Deep Probabilistic Modeling for Causal Inference and Decision Making
- 발행사항
- [Sl] : University of California, Berkeley, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 106 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-01, Section: B.
- 주기사항
- Advisor: Hubbard, Alan Edward.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Berkeley, 2025.
- 초록/해제
- 요약In my dissertation, I propose novel deep probabilistic modeling approaches in various subfields of causal inference and decision making, specifically, true counterfactual inference and sequential design of graph objects.Estimating an individual's counterfactual outcomes under interventions is a challenging task for traditional causal inference and supervised learning approaches when the outcome is high-dimensional (e.g. gene expressions, facial images) and covariates are relatively limited. In this case, to predict one's outcomes under counterfactual treatments, it is crucial to leverage individual information contained in the observed outcome in addition to the covariates. Prior works using variational inference in counterfactual generative modeling have been focusing on neural adaptations and model variants within the conditional variational autoencoder formulation, which I argue is fundamentally ill-suited to the notion of counterfactual in causal inference. In this work, I present a novel variational Bayesian causal inference framework and its theoretical backings to properly handle counterfactual generative modeling tasks, through which we are able to conduct counterfactual supervision end-to-end during training without any counterfactual samples, and encourage disentangled exogenous noise abduction that aids the correct identification of causal effect in counterfactual generations. In addition, I present the efficient influence function and a scheme for robust marginal and heterogeneous effect estimation within this framework. In experiments, I demonstrate the advantage of this framework compared to state-of-the-art models in counterfactual generative modeling on multiple benchmarks.In particular, predicting counterfactual responses of a cell under drug/gene perturbations may bring important benefits to drug discovery and personalized therapeutics. In this work, I present a novel graph variational Bayesian causal inference framework to predict a cell's gene expressions under counterfactual perturbations (perturbations that this cell did not factually receive), leveraging information representing biological knowledge in the form of gene regulatory networks (GRNs) to aid individualized cellular response predictions. Aiming at a data-adaptive GRN, I also present an adjacency matrix updating technique for graph convolutional networks and used it to refine GRNs during pre-training, which generated more insights on gene relations and enhanced model performance. With extensive experiments, I exhibited the advantage of this approach over state-of-the-art deep learning models for individual response prediction.In addition, to directly tackle drug discovery tasks, I propose Distilled Graph Attention Policy Network (DGAPN), a reinforcement learning model to generate novel graph-structured chemical representations that optimize user-defined objectives by efficiently navigating a physically constrained domain. The framework is examined on the task of generating molecules that are designed to bind, noncovalently, to functional sites of SARS-CoV-2 proteins. I present a spatial Graph Attention (sGAT) mechanism that leverages self-attention over both node and edge attributes as well as encoding the spatial structure - this capability is of considerable interest in synthetic biology and drug discovery. An attentional policy network is introduced to learn the decision rules for a dynamic, fragment-based chemical environment, and state-of-the-art policy gradient techniques are employed to train the network with stability. Exploration is driven by the stochasticity of the action space design and the innovation reward bonuses learned and proposed by random network distillation. In experiments, this framework achieved outstanding results compared to state-of-the-art algorithms, while reducing the complexity of paths to chemical synthesis.
- 일반주제명
- Biostatistics
- 일반주제명
- Computer engineering
- 일반주제명
- Computer science
- 키워드
- Causal inference
- 키워드
- Decision making
- 기타저자
- University of California, Berkeley Biostatistics
- 기본자료저록
- Dissertations Abstracts International. 87-01B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017357021
■00520260202103118
■006m o d
■007cr#unu||||||||
■020 ▼a9798288861840
■035 ▼a(MiAaPQ)AAI31937498
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aWu, Yulun.
■24510▼aDeep Probabilistic Modeling for Causal Inference and Decision Making
■260 ▼a[Sl]▼bUniversity of California, Berkeley▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a106 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-01, Section: B.
■500 ▼aAdvisor: Hubbard, Alan Edward.
■5021 ▼aThesis (Ph.D.)--University of California, Berkeley, 2025.
■520 ▼aIn my dissertation, I propose novel deep probabilistic modeling approaches in various subfields of causal inference and decision making, specifically, true counterfactual inference and sequential design of graph objects.Estimating an individual's counterfactual outcomes under interventions is a challenging task for traditional causal inference and supervised learning approaches when the outcome is high-dimensional (e.g. gene expressions, facial images) and covariates are relatively limited. In this case, to predict one's outcomes under counterfactual treatments, it is crucial to leverage individual information contained in the observed outcome in addition to the covariates. Prior works using variational inference in counterfactual generative modeling have been focusing on neural adaptations and model variants within the conditional variational autoencoder formulation, which I argue is fundamentally ill-suited to the notion of counterfactual in causal inference. In this work, I present a novel variational Bayesian causal inference framework and its theoretical backings to properly handle counterfactual generative modeling tasks, through which we are able to conduct counterfactual supervision end-to-end during training without any counterfactual samples, and encourage disentangled exogenous noise abduction that aids the correct identification of causal effect in counterfactual generations. In addition, I present the efficient influence function and a scheme for robust marginal and heterogeneous effect estimation within this framework. In experiments, I demonstrate the advantage of this framework compared to state-of-the-art models in counterfactual generative modeling on multiple benchmarks.In particular, predicting counterfactual responses of a cell under drug/gene perturbations may bring important benefits to drug discovery and personalized therapeutics. In this work, I present a novel graph variational Bayesian causal inference framework to predict a cell's gene expressions under counterfactual perturbations (perturbations that this cell did not factually receive), leveraging information representing biological knowledge in the form of gene regulatory networks (GRNs) to aid individualized cellular response predictions. Aiming at a data-adaptive GRN, I also present an adjacency matrix updating technique for graph convolutional networks and used it to refine GRNs during pre-training, which generated more insights on gene relations and enhanced model performance. With extensive experiments, I exhibited the advantage of this approach over state-of-the-art deep learning models for individual response prediction.In addition, to directly tackle drug discovery tasks, I propose Distilled Graph Attention Policy Network (DGAPN), a reinforcement learning model to generate novel graph-structured chemical representations that optimize user-defined objectives by efficiently navigating a physically constrained domain. The framework is examined on the task of generating molecules that are designed to bind, noncovalently, to functional sites of SARS-CoV-2 proteins. I present a spatial Graph Attention (sGAT) mechanism that leverages self-attention over both node and edge attributes as well as encoding the spatial structure - this capability is of considerable interest in synthetic biology and drug discovery. An attentional policy network is introduced to learn the decision rules for a dynamic, fragment-based chemical environment, and state-of-the-art policy gradient techniques are employed to train the network with stability. Exploration is driven by the stochasticity of the action space design and the innovation reward bonuses learned and proposed by random network distillation. In experiments, this framework achieved outstanding results compared to state-of-the-art algorithms, while reducing the complexity of paths to chemical synthesis.
■590 ▼aSchool code: 0028.
■650 4▼aBiostatistics
■650 4▼aComputer engineering
■650 4▼aComputer science
■653 ▼aCausal inference
■653 ▼aCounterfactual inference
■653 ▼aDecision making
■653 ▼aDeep probabilistic modeling
■653 ▼aGraph neural networks
■653 ▼aReinforcement learning
■690 ▼a0308
■690 ▼a0984
■690 ▼a0464
■71020▼aUniversity of California, Berkeley▼bBiostatistics.
■7730 ▼tDissertations Abstracts International▼g87-01B.
■790 ▼a0028
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17357021▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


