Designing minimal sorting networks using a bio-inspired technique

Sorting Networks (SN) are efficient tools to sort an input data sequence. They are composed by a set of comparison-exchange operations called comparators. The comparators are a priori fixed for a determined input size. The comparators are independent of the input configuration. SN with a minimal number of comparators results in an optimal manner to sort data; it is a classical NP-hard problem studied for more than 50 years. In this paper we adapted a biological inspired heuristic called Artificial Immune System to evolve candidate sets of SN. Besides, a local strategy is proposed to consider the information regarding comparators and sequences to be ordered at a determined building stage. New optimal Sorting Networks designs for input sizes from 9 to 15 are presented.