Zipper-based Meta-Genetic Programming
Kyle Harrington, Jordan B. Pollack · 2011
We present a zipper-based instruction set for constructing genetic pro-gramming variation operators. We study the effects of such variation op-erators on the lawnmower problem. Operators in this language possess the ability to outperform standard mutation and crossover in terms of both decreasing the average population error as well as the best population error. Furthermore, the expression of standard mutation and crossover in this language is trivial, allowing evolution to re-discover these operators when necessary. We conclude by using these operators in a meta-genetic programming context, showing that it is possible for zipper-based meta-genetic programming to expedite evolutionary search. 1