Radioengineering (Apr 2013)

Multi-Objective Self-Organizing Migrating Algorithm: Sensitivity on Controlling Parameters

  • P. Kadlec,
  • Z. Raida,
  • J. Drinovsky

Journal volume & issue
Vol. 22, no. 1
pp. 296 – 308

Abstract

Read online

In this paper, we investigate the sensitivity of a novel Multi-Objective Self-Organizing Migrating Algorithm (MOSOMA) on setting its control parameters. Usually, efficiency and accuracy of searching for a solution depends on the settings of a used stochastic algorithm, because multi-objective optimization problems are highly non-linear. In the paper, the sensitivity analysis is performed exploiting a large number of benchmark problems having different properties (the number of optimized parameters, the shape of a Pareto front, etc.). The quality of solutions revealed by MOSOMA is evaluated in terms of a generational distance, a spread and a hyper-volume error. Recommendations for proper settings of the algorithm are derived: These recommendations should help a user to set the algorithm for any multi-objective task without prior knowledge about the solved problem.

Keywords