CAREER: Robustness, Active Learning, Sparsity, and Fairness in Classification

职业：分类中的鲁棒性、主动学习、稀疏性和公平性

• 批准号: 2239376
• 负责人: Jie Shen
• 金额: $ 59.07万
• 依托单位: Stevens Institute of Technology
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-07-01 至 2028-06-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2239376&HistoricalAwards=false
• 关键词: CAREER; Robustness; Active; Learning; Sparsity

Learning and making accurate inferences from complex data is a core task in modern data analytics. Over the past few decades, a large volume of efficient algorithms has been developed and tested in a broad range of science and engineering problems. However, rigorous analysis of these algorithms often relies on simplified assumptions about the structure of the data, which may not capture the real characteristics. The project aims to go beyond standard theoretical frameworks by developing new theories and algorithms to address pressing challenges arising from contemporary applications, such as adversarial data contamination. In particular, this project will study and lay solid theoretical foundations for the classification problem, which plays a fundamental role in machine learning. A crucial educational component of the project involves the development of a new undergraduate program in artificial intelligence and machine learning that has the potential to inspire a transformation of nationwide STEM education. Furthermore, the principal investigator will continue to mentor undergraduate and graduate students.The project will address several fundamental questions in classification for which there is a large gap in our current understanding. A wide range of modern tools will be leveraged to design new algorithms that can tolerate adversarial corruptions in the data, mitigate data annotation costs, circumvent the curse of high dimensionality, and fortify models with fairness guarantees. Complementary to the algorithmic results, the project will also develop information-theoretic and statistical-query lower bounds to broaden the understanding of fundamental limits posed by the practical constraints. The comprehensive investigation of these problems and their interplay will lead to new analytic and algorithmic tools, enrich various areas (such as learning theory, statistics, and optimization), and build new bridges between them.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.

从复杂数据中学习和做出准确的推断是现代数据分析的核心任务。在过去的几十年里，大量的高效算法已经被开发出来，并在广泛的科学和工程问题中进行了测试。然而，对这些算法的严格分析通常依赖于对数据结构的简化假设，这可能无法捕获真实的特征。该项目旨在通过开发新的理论和算法来超越标准的理论框架，以解决当代应用中出现的紧迫挑战，例如对抗性数据污染。特别是，本项目将研究并奠定坚实的理论基础，分类问题，这在机器学习中起着基础性的作用。该项目的一个重要教育组成部分涉及开发一个新的人工智能和机器学习本科课程，该课程有可能激发全国STEM教育的转型。此外，首席研究员将继续指导本科生和研究生。该项目将解决分类中的几个基本问题，这些问题在我们目前的理解中存在很大差距。将利用各种现代工具来设计新的算法，这些算法可以容忍数据中的对抗性破坏，降低数据注释成本，规避高维灾难，并通过公平性保证来强化模型。作为对算法结果的补充，该项目还将开发信息理论和实践查询下限，以扩大对实际约束所造成的基本限制的理解。对这些问题及其相互作用的全面调查将导致新的分析和算法工具，丰富各个领域（如学习理论，统计和优化），并在它们之间建立新的桥梁。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。