Many applications in data management can benefit from a “data sketch”: a compact data structure that captures the important features of a much larger object. Over the years, many different sketches have been proposed for finding a variety of foundational statistics about an evolving dataset: the number of distinct elements, the median element, the frequently occurring elements (heavy hitters), and several other metrics and norms. These have been used by the data management community for a number of purposes: for efficient measurements when the full data is too large to retain; to guide query planning; and to support approximate query answering.
This file was generated by bibtex2html 1.92.