CDS&E: AAG: "glupyter": Enabling multi-dimensional linked data visualization with glue in the browser

开源glue包允许科学家探索相关数据集内部和之间的关系。它可以很容易地创建多维链接可视化，以交互方式或编程方式选择子集，然后在上下文中查看这些选择。glue的一个独特特性是，它允许用户在数据组件集之间建立用户定义的数学关系。由于目前胶水应用的局限性，一种新的胶水接口正在开发中，称为“glupyter”。目前的奖项将支持将glupyter从一个概念验证应用程序转变为一个功能齐全、完善和科学级的应用程序，用于天文研究和其他领域。Glue和glupyter可以用于许多科学领域，从探索天文观测到分析大脑的放射图像，再到探索多波长卫星图像中的模式。随着它们能力的扩大，它们在科学研究之外的潜在影响也将扩大。它们已经是教授数据探索基础知识的好工具，并将为6-14年级的学生提供新的数据科学课程。认知和可视化研究早就表明，人类特别擅长观察变化，这就是为什么使用高维数据的链接视图进行实时数据探索，即使在非常大的数据集中，也能提供如此有效的方法来寻找趋势、异常值和显著子集。商业链接视图工具可以用于单个表格数据文件，但是现有的工具无法处理跨多个文件的二维图像和三维卷中的链接视图选择。这就是胶水的强度所在。同时，Jupyter笔记本在数据分析方面越来越受欢迎，因为它们使用叙述性文本、代码和可视化来存储整个工作流。它们鼓励可复制性并促进共享。将glue带入木星生态系统作为glupyter提供了无与伦比的洞察力潜力。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。

The open-source glue package allows scientists to explore relationships within and across related datasets. It makes it easy to create multi-dimensional linked visualizations, to select subsets interactively or programmatically, and then see those selections in context. A unique feature of glue is that it allows user-defined mathematical relationships between sets of data components. Because of limitations with the current glue application, a new interface to glue is under development, called "glupyter". The present award will support transforming glupyter from a proof-of-concept application to a fully featured, polished and science-grade application for astronomical research and beyond. Glue and glupyter can be used across many fields of science, from exploring astronomical observations to analyzing radiological images of the brain to exploring patterns in multi-wavelength satellite images. As their capabilities expand, so will their potential for impact outside of scientific research. They are already good tools for teaching the fundamentals of data exploration, and will enable new data-science curricula for students in grades 6-14. Cognitive and visualization research has long shown that humans are especially adept at seeing change, which is why real-time data exploration using linked views of high-dimensional data offers such an effective way to look for trends, outliers and salient subsets, even in very large data sets. Commercial linked-view tools can be used for single tabular data files, but no extant tool handles linked-view selection in two-dimensional images and three-dimensional volumes spread across multiple files. This is where the strength of glue lies. Meanwhile, Jupyter notebooks are increasingly popular for data analysis, as they store an entire workflow using narrative text, code, and visualizations. They encourage reproducibility and facilitate sharing. Bringing glue into the Jupyter ecosystem as glupyter offers unparalleled potential for insight.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.