Sensitivity Analysis Applied to Decision Making in Multiobjective Evolutionary Optimization

Sergio Luciano Avila (1, 2) , A.C. Lisboa (3) , Laurent Krähenbühl (1) , Walter P. Carpes (2) , Joao Antonio Vasconcelos (3) , R. R. Saldanha (3) , Ricardo H. C. Takahashi (4)

1 CEGELY - Centre de génie électrique de Lyon
2 GRUCAD - Grupo de concepção e analise de dispositivos eletromagnéticos
3 DEE - UFMG - Departamento de Engenharia Elétrica [Minas Gerais]
4 DM - UFMG - Departamento de Matemática [Minas Gerais]

The result of a multiobjective evolutionary optimization is an efficient solution set surrounded by other candidate solution points. To choose a final solution, we can perform a sensitivity study. Applying this methodology, disturbances that occur in real-world design problems are not neglected. This paper presents an easy way to perform the sensitivity analysis directly from the data generated from a multiobjective stochastic optimization process. No additional function evaluation is required. As an example, we have solved some optimization problems concerning electromagnetic devices.

Mots clés: Multiobjective evolutionary optimization; sensitivity analysis

Domaines: Autre

Sergio Luciano Avila, A.C. Lisboa, Laurent Krähenbühl, Walter P. Carpes, Joao Antonio Vasconcelos, et al.. Sensitivity Analysis Applied to Decision Making in Multiobjective Evolutionary Optimization. IEEE Transactions on Magnetics, 2006, 42 (4), pp.1103-1106. ⟨10.1109/TMAG.2006.871447⟩. ⟨hal-00111247⟩