Detecting and studying light echoes in the era of Rubin and Artificial Intelligence

该奖项支持建立一个基于人工智能（AI）的管道，用于全天候自动检测和研究光回波。光回波（LEs）是由恒星爆炸点燃宇宙尘埃引起的，它们微弱、罕见、弥漫，很难探测到。对所有天空样本的探测可以发现以前未知的银河系超新星，并允许研究银河系尘埃和银河系中恒星爆炸和喷发的历史。这些新的人工智能方法将在现有和未来的调查数据集中重新发现旧的和新的轻回声。该项目将培养特拉华大学（University of Delaware）和特拉华州立大学（Delaware State University）学生的基本数据科学技能。特拉华州立大学是为少数族裔和农村人口服务的HBCU。用于高效可靠地检测低信噪比漫射特征的人工智能模型架构可用于天文学、医学成像以及生态学和城市代谢研究。目前，视差探测是通过目测来实现的，这是有局限性的，不能适用于全天巡天。Vera C. Rubin天文台遗留时空巡天（LSST）将频繁观测南方天空，使其成为理想的LE巡天。不幸的是，LSST警报管道针对点源进行了优化，将完全错过LEs。该项目将利用尖端的人工智能技术，建立首个用于LEs全天自动探测和研究的管道，并支持该管道在LSST科学平台上的部署。该团队还将为来自STEM领域历史上代表性不足的群体的学生提供动手数据和计算机科学培训，采用沉浸式学习计划，包括数据科学训练营、黑客马拉松和指导研究机会。该项目利用了NSF的两大构想：“利用数据革命”和“日益融合的研究”。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。

This award supports building an artificial intelligence (AI)-based pipeline for the all-sky-scale automated detection and study of light echoes. Light echoes (LEs) are caused by stellar explosions lighting up cosmic dust, and they are faint, rare, diffuse, and hard to detect. The detection of all sky samples of LEs can enable the discovery of previously unknown Galactic supernovae, and permit the study of Galactic dust and the history of stellar explosions and eruptions in the Galaxy. These new AI methods will robustly re-discover old, and discover new, light echoes in both existing and future survey datasets. The project will develop essential data science skills in students at the University of Delaware and at Delaware State University, a minority- and rural population-serving HBCU. AI model architectures for efficient and reliable detection of low-signal-to-noise diffuse features can be used throughout astronomy, for medical imaging, and in ecology and urban metabolism studies.Presently, LEs are detected by visual inspection, which is limiting and does not scale to all-sky surveys. The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will observe the southern sky at frequent intervals, making it an ideal LE survey. Unfortunately, the LSST alert pipeline is optimized for point sources and will entirely miss LEs. This project will produce the first pipeline for the automated all-sky detection and study of LEs by leveraging cutting-edge AI, and support deployment of the pipeline on the LSST science platform. The team will also be providing hands-on data- and computer-science training to students from groups historically underrepresented in the STEM fields, using an immersive learning program that includes Data Science boot camps, hackathons, and mentored research opportunities. This project capitalizes on the two NSF Big Ideas of “Harnessing the Data Revolution” and “Growing Convergence Research”.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.