Multi-target swarm intelligence algorithm parallel optimization method based on cloud computing

Abstract

The invention discloses a multi-target swarm intelligence algorithm parallel optimization method based on cloud computing, and the method comprises the following steps: parallelizing an original standalone version multi-target swarm intelligence optimization method by using a GPU platform in the cloud computing, serially executing cross selection mutation operation on a CPU, and allocating control comparison of population individuals to the GPU for execution, so the time complexity is reduced, and the time required by algorithm execution is reduced. Meanwhile, a multi-island model in the parallel multi-target swarm intelligence algorithm is introduced, simultaneous execution of the multi-target swarm intelligence algorithm with multiple GPU parallelization is achieved, species diversity is increased through a migration strategy, and therefore the solving quality of the algorithm is improved. The multi-island model is specifically realized by adopting a common framework Spark in cloud computing, the number of islands can be increased by modifying parameters, and the method has high expansibility.

Description

technical field [0001] The invention relates to the technical field of distributed and parallel swarm intelligence algorithms, in particular to a multi-objective swarm intelligence algorithm parallel optimization method based on cloud computing. Background technique [0002] The multi-objective optimization problem is a multi-standard decision-making problem. There are multiple conflicting optimization objectives in the problem at the same time, resulting in the absence of a global optimal solution so that all objective functions reach the optimal value at the same time. Multi-objective group intelligent optimization can be used algorithmic solution. The theoretical global optimal solution set is called real Pareto solution, and the solution obtained by evolutionary algorithm is called approximate Pareto solution. The existing multi-objective optimization algorithms can be divided into three categories: Pareto-based multi-objective group intelligence algorithm, indicator-ba...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the time cost of the existing stand-alone version of the multi-objective group intelligent algorithm and the shortcomings that the solution quality is difficult to improve, and proposes a cloud computing-based The multi-objective group intelligent algorithm parallel optimization method first uses the common cloud computing platform GPU to parallelize the stand-alone version of the multi-objective group intelligent algorithm to shorten the running time of the algorithm and reduce the time cost, and then designs multiple islands through the common cloud computing platform Spark The model realizes the simultaneous operation of multiple GPU parallelized multi-objective group intelligent algorithms, and further improves the solution quality of the algorithm through the island model

Images