DPE, proposed by Schraudolph and Belew? changes the individual representation scheme dynamically via searching for the "target interval". It uses a GA with multiple restarts. The first run is used to find an interim solution, and the subsequent runs decode the genes as distances (delta values) from the last interim solution. (from RefLeung2002)
In this, as a GA gets near to what seems to be an optimum, then the genome is recoded, and the search space is "zoomed in on".