Code Growth in Genetic Programming
The MIT Press eBooks · 1996
In this paper we examine how quickly the programs generated using genetic programming grow in size. We found that without a constraint mechanism the programs will grow indefinitely regardless of whether or not the growth acts to improve the programs' solutions. This growth is dominated by non-functional code. If the non-functional code is removed the growth is dominated by functional, but non-executed code. Two methods of controlling the growth: removing non-functional code, and selective pressure applied by penalizing longer programs, are compared. Only the later method appears to be effective in bounding the programs' size.