Implementation Models for Distributed Memory Architecture of Parallel Simulated Annealing Using Genetic Crossover

This paper examines implementation models for distributed memory architectures of a Parallel Simulated Annealing using Genetic Crossover (PSA/GAc). The PSA/GAc that was proposed by authors is the algorithm, where there are several processes of a simulated annealing working parallel. To exchange information between the solutions, the operation of genetic crossover is performed. We need new models to implement PSA/GAc to distributed memory architecture such as a PC cluster system, since PSA/GAc was designed only for shared memory architecture. We developed three types of implementation models of PSA/GAc. Each model was applied to a protein structure prediction problem that is one of the optimization problems. This paper makes a comparison and examination the effectiveness between the proposed models from two points of view; those are a computation time and a searching ability. Then, it is found that one of the proposed models are superior to the other models, since it can get more speed up and has high searching ability.