Structure abstraction and genetic programming
Tina Yu · 2003
The selection of program representation can have strong impact on the performance of genetic programming. Previous work has shown that a particular program representation which supports structure abstraction is very effective in solving the general even parity problem. We investigate program structures and analyze all perfect solutions in the search space to provide explanation of why structure abstraction is so effective with this problem. This work provides guidelines for the application of structure abstraction to other problems.