CAREER: Toward A Knowledge-Guided Framework for Personalized Decision Making

职业：走向个性化决策的知识引导框架

• 批准号: 2144209
• 负责人: Jundong Li
• 金额: $ 60万
• 依托单位: University of Virginia Main Campus
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-01 至 2027-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2144209&HistoricalAwards=false
• 关键词: CAREER; Toward; Knowledge; Guided; Framework

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).Learning causality from data is a vital stepping stone toward building human-level intelligent systems that can make appropriate decisions. In seeking to make an optimal decision for each individual (i.e., personalized decision making), we need to understand the causal relationship between a decision and its consequent outcome. Causal inference provides a principled way to achieve personalized decision making by learning individual-level causal effects from observational data. Its impacts are seen in a broad spectrum of application domains. However, existing causal inference frameworks are mostly data-driven and face multifaceted challenges (at the assumption-, data-, and application-level) when applied in real-world observational studies. Despite that, a vast amount of prior human knowledge manifests itself in different ways and could be leveraged to tackle these challenges. Although abundant human knowledge provides great opportunities, its complex nature coupled with observational data also imposes tremendous hurdles. This project aims to bridge the gap between what can be accessed (i.e., a large amount of observational data across different domains and human knowledge in different formats) and what is desired (i.e., more effective causal inference to advance personalized decision making).This project develops a suite of novel causal inference models and algorithms to analyze observational data by harnessing the power of human knowledge and gaining deeper insights to advance personalized decision making. First, it leverages relational knowledge that describes the relations among data instances in observational data, investigates its role in relaxing overly optimistic assumptions for causal inference. Second, it explores meta knowledge that depicts distinct properties of observational data and develops principled causal inference models and algorithms to incorporate such knowledge. Third, it aims to improve the utility of existing data-driven causal inference frameworks by harnessing application knowledge, which characterizes the unique needs of real-world applications. The outcomes of this project will enable researchers and practitioners to assimilate massive amounts of observational data, across numerous application domains, and leverage abundant human knowledge, to benefit scientific discovery and informed decision making. Outcomes of this project will be integrated into the existing curricula and new courses. This project will also provide research opportunities to undergraduate and graduate students, especially female and underrepresented minorities. Customized research and teaching components will be designed and implemented to attract K-12 students in STEM education and engage them in causal inference and data science research. Last but not least, this project will improve student success and retention via a unique educational decision making component. This approach will optimize current education systems, for the benefit of generations of students to come.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.

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。从数据中学习因果关系是构建能够做出适当决策的人类智能系统的重要基石。在寻求为每个个体做出最佳决策（即个性化决策）时，我们需要了解决策与其结果之间的因果关系。因果推理提供了一种原则性的方法，通过从观察数据中学习个人层面的因果效应来实现个性化决策。它的影响可以在广泛的应用领域中看到。然而，现有的因果推理框架大多是数据驱动的，在应用于现实世界的观察研究时面临多方面的挑战（在假设、数据和应用层面）。尽管如此，大量先前的人类知识以不同的方式表现出来，可以用来应对这些挑战。尽管丰富的人类知识提供了巨大的机会，但其复杂性加上观测数据也带来了巨大的障碍。该项目旨在弥合可访问的内容（即，跨越不同领域的大量观测数据和不同格式的人类知识）与期望的内容（即，更有效的因果推理以推进个性化决策）之间的差距。该项目开发了一套新颖的因果推理模型和算法，通过利用人类知识的力量来分析观测数据，并获得更深入的见解，以推进个性化决策。首先，它利用描述观测数据中数据实例之间关系的关系知识，研究其在放松过度乐观的因果推理假设中的作用。其次，它探索描述观测数据的不同属性的元知识，并开发原则性的因果推理模型和算法来整合这些知识。第三，它旨在通过利用应用知识来提高现有数据驱动的因果推理框架的效用，这些应用知识表征了现实世界应用的独特需求。该项目的成果将使研究人员和从业人员能够吸收大量观测数据，跨越众多应用领域，并利用丰富的人类知识，有利于科学发现和知情决策。该项目的成果将整合到现有课程和新课程中。该项目还将为本科生和研究生，特别是女性和代表性不足的少数民族提供研究机会。定制的研究和教学组件将被设计和实施，以吸引K-12学生参与STEM教育，并让他们参与因果推理和数据科学研究。最后但并非最不重要的是，这个项目将通过一个独特的教育决策组件来提高学生的成功和保留率。这种方法将优化当前的教育系统，造福未来几代学生。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。

项目成果

Federated Few-shot Learning

• DOI: 10.1145/3580305.3599347
• 发表时间: 2023-06
• 期刊: Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
• 作者: Song Wang;Xingbo Fu;Kaize Ding;Chen Chen-Chen;Huiyuan Chen;Jundong Li

Fairness in Graph Mining: A Survey

• DOI: 10.1109/tkde.2023.3265598
• 发表时间: 2022-04
• 期刊: IEEE Transactions on Knowledge and Data Engineering
• 影响因子: 8.9
• 作者: Yushun Dong;Jing Ma;Song Wang;Chen Chen-Chen;Jundong Li

Interpreting Unfairness in Graph Neural Networks via Training Node Attribution

通过训练节点归因解释图神经网络中的不公平性

• DOI: 10.1609/aaai.v37i6.25905
• 发表时间: 2023
• 期刊: Proceedings of the AAAI Conference on Artificial Intelligence
• 作者: Dong, Yushun;Wang, Song;Ma, Jing;Liu, Ninghao;Li, Jundong
• 通讯作者: Li, Jundong

Few-shot Node Classification with Extremely Weak Supervision

• DOI: 10.1145/3539597.3570435
• 发表时间: 2023-01
• 期刊: Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining
• 作者: Song Wang;Yushun Dong;Kaize Ding;Chen Chen-Chen;Jundong Li

Transductive Linear Probing: A Novel Framework for Few-Shot Node Classification

• DOI: 10.48550/arxiv.2212.05606
• 发表时间: 2022-12
• 作者: Zhen Tan;Song Wang;Kaize Ding;Jundong Li;Huan Liu