VFF: Seeing cluster constellations in data point clouds

Exploratory discrimination and classification is often the first task in data analysis, leading to new probes and discoveries. We present unsupervised cluster results, by our new classification algorithm named sparse dual of the density peaks (SD-DP), on several data sets, including benchmark data for testing and real-world data for exploratory investigation. The real-world data include color images, cells via gene expressions, English words in vector representation, and scientific articles. The latter data are collected over the Internet with KIWI, a tool we developed for theme-specific bibliographic search and survey for novice and expert researchers. We also introduce a couple of perspective ways for visual understanding, interpretation, and appreciation of cluster configurations.