Abstract. This work investigates the parallelization of the Artificial Bee Colony Algorithm. Besides a sequential version enhanced with local search, we compare three parallel models: master-slave, multi-hive with migrations, and hybrid hierarchical. Extensive experiments were done using three numerical benchmark functions with a high number of variables. Statistical results indicate that intensive local search improves the quality of solutions found and, thanks to the coevolution effect, the multi-population approaches obtain better quality with less computational effort. A final comparison between models was done analyzing the trade-offs between quality of solution and processing time.

No abstract

This work presents a new evolutionary algorithm based on the standard harmony search strategy, called population-based harmony search (PBHS). Also, this work provides a parallelisation method for the proposed PBHS by using graphical processing units (GPU), allowing multiple function evaluations at the same time. Experiments were done using a benchmark of a hard scientific problem: protein structure prediction with the AB-2D off-lattice model. The performance and the solution quality were evaluated and compared using four implementations: two concerning the standard HS, one running in CPU and another running in GPU, and two implementations concerning the PBHS, also running in CPU and in GPU. Results show that the quality of solutions and speed-ups achieved by the PBHS is significantly better than the HS.

No abstract

This work presents a master-slave parallel genetic algorithm for the protein folding problem, using the 3D-HP side-chain model (3D-HP-SC). This model is sparsely studied in the literature, although more expressive than other lattice models. The fitness function proposed includes information not only about the free-energy of the conformation, but also compactness of the side-chains. Since there is no benchmark available to date for this model, a set of 15 sequences was used, based on a simpler model. Results show that the parallel GA achieved a good level of efficiency and obtained biologically coherent results, suggesting the adequacy of the methodology. Future work will include new biologicallyinspired genetic operators and more experiments to create new benchmarks.