Rethinking the Data-driven Discovery of Rare Phenomena

这个项目将研究数据驱动的罕见现象发现的新技术。科学学科越来越有能力产生与关键发现相关的大量数据，如新型光伏材料或对大脑癫痫的解释。然而，这些发现通常对应于高维空间中极其罕见的现象，而目前的数据科学方法无法检测到这些现象。该项目将填补这一空白，并为数据驱动的发现或罕见现象产生新的方法、出版物和开源软件。因此，它将扩展数据科学的能力，更好地利用科学、政府和行业积累的海量数据集合。

This project will investigate novel technologies for the data-driven discovery of rare phenomena. Scientific disciplines are increasingly able to generate large amounts of data relevant to key discoveries such as novel photovoltaic materials or explanations of brain seizures. However, these discoveries typically correspond to extremely rare phenomena in high dimensional spaces, which current data science methods are unable to detect. The project will fill this void and yield novel methods, publications, and open source software for the data-driven discovery or rare phenomena. Thus, it will expand the capabilities of data science, providing better use of the massive data collections accumulating across science, government, and industry.