In this work, we investigate the ability of a Chunking GA (ChGA) to reduce the size of variable length chromosomes and control bloat. The ChGA consists of a standard genetic algorithm augmented by a communal building block memory system and associated memory chromosomes and operators. A new mxn MaxSum fitness function used for this work is also described. Results show that a ChGA equipped with memory capacity equal to or greater than the minimal size of an optimal solution naturally eliminates unexpressed genes.
Hal Stringer, Annie S. Wu