Using Genetic Programming to Evolve a General Purpose Sorting Network for Comparable Data Sets published in Genetic Algorithms and Genetic Programming at Stanford 2003.
Abstract: This paper demonstrates how non-typed genetic programming may be used to evolve sorting networks; specifically a sorting network for a seven-element integer data sequence. This problem has shown importance over time be- cause programs generated to solve it consume a fixed amount of time and operate in a way that is oblivious to the contents of the data sequence being sorted. Both of these factors lead to sorting networks being more desirable than recursive sorts for data sequences of limited length.
Downloaded the sorting networks paper.