Abstract: |
The Proportional Genetic Algorithm (PGA) supports truly location independent solution encoding. In this paper, we propose a novel approach, called incremental building blocks, to improve the search performance of PGA. The main idea of this approach is to evolve low level building blocks individually in the beginning of a run and then continuously select and combine good building blocks to form larger and more complete ones until an optimal solution is found. Empirical studies in the resource allocation domain show that this approach is able to improve both the search quality and efficiency of a PGA. Further analysis of experimental data reveals that this approach is better at preserving the self-similarity of chromosomes during evolution. |
