Conference: UCLA Synthetic Data Workshop

该奖项支持来自不同数学学科和计算机科学的专家，特别是研究生和其他早期职业研究人员参加即将于2023年4月13日至4月14日在洛杉矶加州大学举行的UCLA合成数据研讨会。研讨会的目标是促进研究人员在与合成数据和数据隐私相关的几个领域的合作，包括差异隐私，公平性和对抗鲁棒性。这项活动的理由是，合成数据生成是一个迅速增长和高度学科化的研究领域，引起了人们的广泛关注。对于欺诈欺骗和垃圾邮件识别的算法程序的开发，以及制造和供应链管理中人工智能驱动模型的构建，合成数据已成为宝贵的资源。本次研讨会的目标是调查这些进步所产生的科学基础，并研究解决开放问题的新策略。讲习班还将有大量的教学内容，以介绍性谈话的形式，涵盖其重点领域的背景和最近令人兴奋的进展。这些讲座将面向非专家，包括研究生和初级研究人员。当获取真实世界的数据成本太高或风险太大时，合成数据尤其有用。最近的结果暗示了一个新的和有希望的方向，即从业者可以通过在使用合成数据的同时解决边缘场景和危险事件来有效地训练AI模型。尽管合成数据的许多成功应用，其科学基础，例如，在保真度、实用性和隐私性之间的权衡仍然缺失。此外，尚未制定生成和利用合成数据的行业标准以及有关合成数据的隐私法。本次研讨会将为专家们提供一个环境，就有关合成数据的开放性问题交换意见，例如在创建合成数据时是否会失去隐私，使用合成数据是否会影响公平性，以及如何在最基本的层面上判断合成数据的质量和有用性。该研讨会的网站是https://ucla-synthetic-data.github.io/.This奖反映了NSF的法定使命，并已被认为值得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的支持。

This award supports participation by experts from diverse mathematical disciplines and computer science, especially graduate students and other early-career researchers, in the upcoming UCLA Synthetic Data Workshop to be held at the University of California, Los Angeles, from April 13 to April 14, 2023. The goal of the workshop is to foster the collaboration of researchers in several areas connected to synthetic data and data privacy, including differential privacy, fairness, and adversarial robustness. The rationale for this activity is that synthetic data generation is a rapidly growing and highly disciplinary research area that draws much attention. For the development of algorithmic procedures for fraud deception and spam identification, as well as for the construction of AI-driven models in manufacturing and supply chain management, synthetic data has become a valuable resource. The goal of this workshop is to investigate scientific foundations that are spawned by these advancements and examine new strategies for solving open problems. The workshop will also have a substantial pedagogical component in the form of introductory talks that will cover background and recent exciting progress in its focus areas. These talks will be accessible to non-experts, including graduate students and junior researchers.Synthetic data is especially useful when obtaining real-world data is either too costly or too risky. Recent results hint at a new and promising direction that practitioners may effectively train AI models by addressing edge scenarios and dangerous occurrences while using synthetic data. Despite numerous successful applications of synthetic data, its scientific foundation, e.g., the tradeoff among fidelity, utility, and privacy, is still missing. In addition, industrial standards for generating and utilizing synthetic data, as well as the privacy law concerning synthetic data, are yet to be established. This workshop will provide an environment for experts to exchange their ideas for open questions about synthetic data, such as whether or not privacy is lost when creating synthetic data, whether or not using synthetic data affects fairness, and how, at the most basic level, one should judge the quality and usefulness of synthetic data. The website for the workshop is https://ucla-synthetic-data.github.io/.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.