Using Choquet integral as preference model in interactive evolutionary multiobjective optimization

abstract We present an interactive evolutionary multiobjective optimization method that can be applied to both continuous and combinatorial problems. The search is driven by user’s preference information elicited in form of pairwise comparisons of some non-dominated solutions in successive generations. This information is then used to curb the set of compatible user’s value functions. When choosing the mathematical form of the preference model guiding the search, one faces the usual dilemma: if the preference model is too simplistic (say, linear), it is unlikely to be able to represent adequately the user’s preferences expressed through the pairwise comparisons; on the other hand, if the preference model is too versatile, a lot of preference information is required 1 

• Type: Journal
• Year: 2016
• Where: EOR
• Authors: Jürgen Branke, Salvatore Corrente, Salvatore Greco, Roman Slowinski, Piotr Zielniewicz
• Keywords: EOR 2016