III: Small: Statistical Inference through Data-Collection and Expert-Knowledge Incorporation

最近的技术进步为研究科学和工程中的各种复杂系统提供了机会，包括生物和环境过程、网络物理系统和社会系统。这些系统的准确建模可以帮助描述它们的时间行为，诊断和预测异常，并预测和控制它们的未来行为。例如，单细胞高通量技术的进步最近为深入了解基因组系统提供了机会，并为慢性疾病的早期诊断、预后和治疗发现了新的有效解决方案。然而，缺乏对目标数据的访问以及获取此类数据的困难对准确建模复杂系统构成了重大挑战。该项目旨在开发算法，以推进数据收集，并将用户和专家知识纳入建模过程。此外，该研究将支持东北大学多样化的博士和本科生群体的跨学科发展，并通过多种举措促进研究的多样性。该项目旨在通过开发利用强化学习、贝叶斯统计和机器学习等先进技术的统计方法来推进动态系统的推理。该项目的技术目标分为两个重点。第一个重点是发展统计学习技术，有效地捕获和量化专家对推理过程的有价值的见解。这些方法量化了数据中包含的专家知识，而不需要专家监督。第二个推力发展了系统化数据收集的方法，以便对复杂系统和过程进行因果和准确的推断。建议的方法利用系统和数据结构，以及相关的不确定性，以实现有效的数据收集。该项目将提出的方法应用于基因组学，包括基因调控网络的拓扑推断。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Recent technological advances have provided opportunities for studying a wide range of complex systems in science and engineering, including biological and environmental processes, cyber-physical systems, and social systems. Accurate modeling of these systems can help characterize their temporal behavior, diagnose and forecast abnormalities, and predict and control their future behavior. For instance, advances in single-cell high-throughput technology have recently provided the opportunity for the deep system-level understanding of genomic systems and the discovery of new effective solutions for early diagnosis, prognosis, and treatment of chronic diseases. However, the lack of access to targeted data and the difficulty of acquiring such data pose significant challenges to accurately modeling complex systems. This project aims to develop algorithms that advance data collection and incorporate user and expert knowledge into the modeling process. Furthermore, the research will support the cross-disciplinary development of a diverse cohort of PhD and undergraduate students at Northeastern University and promote diversity in research through multiple initiatives. This project aims to advance inference of dynamical systems by developing statistical methods that leverage advanced techniques such as reinforcement learning, Bayesian statistics, and machine learning. The technical aims of the project are divided into two thrusts. The first thrust develops statistical learning techniques that effectively capture and quantify valuable expert insights into the inference process. These methods quantify the expert knowledge incorporated within data without requiring expert oversight. The second thrust develops approaches to systematize data collection for causal and accurate inference of complex systems and processes. The proposed approaches leverage the system and data structures, as well as the associated uncertainty, to achieve effective data collection. The project employs the proposed methods in genomics applications, including topology inference in gene regulatory networks.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.