Answers are grounded in this paper and cite their sources.
In Hoare's (1961) original version of the algorithm the partitioning element in the central divide-and-conquer step is chosen uniformly at random from the set S in question. Here we consider a variant where this element is the median of a sample of size 2k+1 from S. We investigate convergence in distribution of the number of comparisons required and obtain a simple explicit result for the limiting average performance of the median-of-three version.
No takes yet. Share an insight, caveat, or question.
Rudolf Grübel (1999) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: