This paper proposes a new approach to cope with multiobjective optimization problems in presence of noise. In the first place, since considering the worst-case performance is important in many realworld optimization problems, a solution is evaluated based on the upper bounds of respective noisy objective functions predicted statistically by multiple sampling. Secondary, a rational way to decide the maximum sample size for the solution is shown. Thirdly, to allocate the computing budget of a proposed evolutionary algorithm only to promising solutions, two pruning techniques are contrived to judge hopeless solutions only by a few sampling and skip the evaluation of the upper bounds for them.
CITATION STYLE
Tagawa, K., & Harada, S. (2014). Multi-noisy-objective optimization based on prediction of worst-case performance. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8890, pp. 23–34). Springer Verlag. https://doi.org/10.1007/978-3-319-13749-0_3
Mendeley helps you to discover research relevant for your work.