Parallel cEAs - The Most Recent Proposals

Here, the most recent proposals for parallelizing cEAs up to now have been collected. If you develope a new algorithm, please let us know and it will be added to this site.

CAGE - CellulAr GEnetic programming

This model of parallel cEA was proposed by Folino et al. in [FPS03] for the case of the genetic programming. CAGE implements the cellular GP model using a one-dimensional decomposition (in the x direction) of the grid and an explicit message passing to exchange information among the domains.

The authors compare CAGE with both the island and the sequential single-population models, obtaining very good results for the CAGE approach.

Figure 1. An example of the parallel model used in CAGE.

Combined Cellular Genetic Algorithms

In the combined cellular genetic algorithm [NAI02], the cellular population is partitioned into several smaller cellular subpopulations. Only the individuals located at the edges of these subpopulations can interact with the individuals of other subpopulations, wich also must be placed in their edges. In Figure 2 it can be seen an example of how the cellular population is partitioned.

Figure 2. An example of the proposed method (four subpopulations with a 10x10 structure)

In [NAI02], the authors examine the effects of parameter specifications such as the number of the subpopulations, the way of placing elite individuals, and the topology of the subpopulations on the performance of the combined cellular genetic algorithms. With respect to the shape of the subpopulations, the authors found that the algorithm performs better when the subpopulations are square than in the case of being linear (method of Folino et al. (CAGE) when the slices are of size one).