Guo Yu, Yaochu Jin, Markus Olhofer, "References or Preferences – Rethinking Many-objective Evolutionary Optimization", CEC 2019, 2019.

Past decades have witnessed a rapid development in multi- and many-objective evolutionary optimization. The references-assisted and preference-driven strategies are both widely used in dealing with the multi- and many-objective optimization problems. However, few research analyzes the difference between these two strategies. Thus, this paper analyzes and compares both strategies from background, constructions, similarities, differences, to concerns. We have found that on the one hand, the references-assisted optimization need to be robust to deal with irregular problems by embedding the preference information into the optimization to better explore the objective space, so that provide a good representative solution set to the decision maker. On the other hand, without sufficient priori-knowledge, it makes the decision maker difficult to provide an informed preference articulation in the preference-driven optimization. Thus, this paper suggests a new way to deal with the multi- and many-objective optimization problems that is to {\textit{a priori}} find the naturally interested regions of the problems like the knee regions. After that, the acquired knowledge of the knee regions can be used in both strategies to gain a more insightful look at the problem.