Sorting Network Development Using Cellular Automata

The sorting network design represents a task that has often beenconsidered as a benchmark for various applications of evolutionary
designand optimization techniques. Although the specific structure of thisclass of circuits allows to use a simple encoding in combination withadditional mechanisms for optimizing the area- and delay-efficiency ofdesigned sorting networks, the design of large sorting networksrepresents a difficult task. This paper proposes a novel cellularautomaton-based approach for the development of specific instances ofsorting networks. In order to explore the area of generative cellularautomata applied on this specific circuit structures, two dierentencodings are introduced: (1) an absolute encoding and (2) a relativeencoding. The abilities of the both techniques are investigated and acomparative study is provided considering a variety of experimentalsettings.