Testing parallelization paradigms for MOEAs
Association for Computing Machinery, Inc
Computing, Health and Science
School of Computer and Information Science
In this paper, we report on our investigation of factors affecting the performance of various parallelization paradigms for multiobjective evolutionary algorithms. Different parallelization paradigms emphasize separate development of sub-populations versus communication and coordination between sub-populations to greater or lesser degrees. We hypothesized that the characteristics of a particular problem will favour some paradigms over others. We tested this hypothesis by creating variations on test problems with different characteristics, and testing the performance of different paradigms in a cluster environment.