RI: Small: Probabilistic Networks for Automated Reasoning

本研究的最终目的是开发能够在动态和不确定环境中自主操作的计算机系统。该研究包括理论和实验研究，旨在融合概率信息之上的因果和反事实关系，以改善数据密集型应用中的诊断、态势评估和决策任务。需要调查的具体研究问题包括开发用于诊断和情况评估的算法和概念工具，使用因果关系和反事实关系，以及开发使用倾向分数估计和c等效测试的结构学习新方法，并将其应用于流行病学和分子生物学。这项研究的理论部分触及人类知识和科学探究的核心，并在流行病学、社会科学、经济学、医学和生物学等领域产生了深刻的方法论影响，在这些领域，因果知识起着重要作用。实践部分侧重于因果和反事实推理，决策制定和结构学习的新算法的开发，并直接应用于使用图形方法编码知识的领域。

The ultimate aim of this investigation is to develop computer systems capable of operating autonomously in dynamic and uncertain environments. The investigation comprises theoretical and experimental studies aiming at fusing causal and counterfactual relationships on top of probabilistic information to improve tasks of diagnosis, situation assessment and decision making in data-intensive applications. Specific research issues to investigate include developing algorithms and conceptual tools for diagnosis and situation assessment, using causal and counterfactual relationships, and developing novel methods of structure learning using propensity-score estimation and C-equivalence tests, with applications to epidemiology and molecular biology. The theoretical part of this research touches on the core of human knowledge and scientific inquiry, and is having profound methodological ramifications in the fields of epidemiology, social science, economics, medicine and biology, where causal knowledge plays a major role. The practical part focuses on the development of new algorithms for causal and counterfactual reasoning, decision making and structure learning, with direct applications in areas that use graphical methods to encode knowledge.