Watson, R. A. “A study of the Effect of Group Formation on Evolutionary Search.” PhD diss., MSc dissertation, Sussex University, 1996.
URL1 URL2
The formation of mutually beneficial groups in an ecosystem may enable the creation of complex entities that would otherwise be unevolvable. This process, symbiogenesis,1 is believed to have enabled certain of the “major transitions in evolution”, [Maynard-Smith & Szathmary, 1995]. This work describes several computational models which investigate the mechanisms involved in enabling symbiogenesis. The experiments attempt to establish both how pervasive the influence of mutually beneficial groups is in nature, and the applicability of these mechanisms to practical engineering problems through Artificial Evolution. The models include a ‘Group Effect’ which shows how the presence of mutually beneficial groups in an ecosystem can alter the fitness landscape and thereby enable the evolution of more complex species. Another model demonstrates how this Group Effect may enable the re-combination of ‘building-blocks’ in a manner which provides a real computational advantage over genetic re-combination in the standard genetic algorithm. Such mechanisms may enable significant increases in the level of complexity that can be attained in evolution, both natural and artificial.