CAREER: A Foundation for Unsupervised Learning of High-Dimensional Data

这项研究旨在开发理论、算法和交互式可视化工具来挖掘来自现实世界应用领域的各种高维数据，从医学图像到卫星数据。该研究旨在开发一个无监督特征选择的统一框架，识别和表征不同聚类目标的聚类标准，创建反映特定领域(例如地球科学图像)中数据的空间和时间性质的新的相似性度量，定义特征相关性/不相关性的度量，并将特征选择纳入层次聚类方法。开发的新算法将使各种应用领域受益；尤其是，它将直接帮助医生研究肺气肿和囊性纤维性肺部疾病的严重程度，并帮助科学家发现地球表面有趣的模式。该项目通过在课堂和实验室为本科生和研究生提供实践研究体验，将研究和教育融为一体。此外，PI计划与女性工程师协会和Connections计划合作，鼓励女高中生追求工程和计算机科学的职业生涯，并确保代表不足的群体参与这项研究。

This research aims to develop theory, algorithms, and interactive visualization tools for mining a variety of high-dimensional data stemming from real world application areas, ranging from medical images to satellite data. The research aims to develop a unified framework for unsupervised feature selection, identify and characterize clustering criteria for different clustering objectives, create new similarity metrics that reflect the spatial and temporal nature of data in specific domains (e.g., earth science images), define measures of feature relevance/irrelevance, and incorporate feature selection into hierarchical clustering methods. The new algorithms developed will benefit a variety of application domains; in particular, it will directly aid physicians studying the severity of emphysema and cystic fibrosis lung diseases, and help scientists discover interesting patterns of the earth surface. This project integrates research and education by providing hands-on research experiences to both undergraduate and graduate students in the classroom and in the lab. Moreover, the PI plans to work with the Society of Women Engineers and the Connections program to inspire female high school students to pursue careers in engineering and computer science and to insure that under-represented groups are involved in this research.