A Knee-Based EMO Algorithm with an Efficient Method to Update Mobile Reference Points
A number of evolutionary multi-objective optimization (EMO) algorithms have been proposed to search for non-dominated solutions around reference points that are usually assumed to be given by a decision maker (DM) based on his/her preference. However, setting the reference point needs a priori knowledge that the DM sometimes does not have. In order to obtain favorable solutions without a priori knowledge, “knee points” can be used. Some algorithms have already been proposed to obtain solutions around the knee points. TKR-NSGA-II is one of them. In this algorithm, the DM is supposed to specify the number of knee points as a parameter whereas such information is usually unknown. In this paper, we propose an EMO algorithm that does not require the DM to specify the number of knee points in advance. We demonstrate that the proposed method can efficiently find solutions around knee points.
KeywordsKnee point Preference Reference point Decision maker
Unable to display preview. Download preview PDF.
- 1.Bechikh, S., Said, L.B., Ghedira, K.: Searching for knee regions in multi-objecitve optimization using mobile reference points. In: Proc. of the 2010 ACM Symposium on Applied Computing, pp. 1118–1125. ACM (2010)Google Scholar
- 8.Narukawa, K., Tanigaki, Y., Ishibuchi, H.: Evolutionary many-objective optimization using preference on hyperplane. In: Proc. of 2014 Conference Companion on Genetic and Evolutionary Computation Conference, pp. 91–92. ACM (2014)Google Scholar