Parallel and Distributed Implementation Models for Bio-inspired Optimization Algorithms

Bio-inspired optimization algorithms have natural parallelism but practical implementations in parallel and distributed computational systems are nontrivial. Gains from different parallelism philosophies and implementation strategies may vary widely. In this paper, we contribute with a new taxonomy for various parallel and distributed implementation models of metaheuristic optimization. This taxonomy is based on three factors that every parallel and distributed metaheuristic implementation needs to consider: control, data, and memory. According to our taxonomy, we categorize different parallel and distributed bio-inspired models as well as local search metaheuristic models. We also introduce a new designed GPU parallel model for the Kohonen's self-organizing map, as a representative example which belongs to a significant category in our taxonomy.