Parallel Feature Selection Using Only Counts

  • Staal Vinterbo
  • Jialan Que

Abstract

Count queries belong to a class of summary statistics routinely used in basket analysis, inventory tracking, and study cohort finding. In this article, we demonstrate how it is possible to use simple count queries for parallelizing sequential data mining algorithms. Specifically, we parallelize a published algorithm for finding minimum sets of discriminating features and demonstrate that the parallel speedup is close to the expected optimum. 

Published
2018-08-08
Section
Artikler