@inproceedings {pub2799,
	title = {An Adaptive Reference Vector Generation Strategy for Inverse Model Based Evolutionary Multiobjective Optimization},
	author = {Ran Cheng AND Yaochu Jin AND Kaname Narukawa},
	year = {2015},
	month = {March},
	abstract = {The objective space and the decision space are two spaces in multiobjective optimization problems (MOPs). Very recently, noticing the fact that to maintain diversity in the objective space is much easier than in the decision space, the authors have proposed an inverse modeling based multiobjective evolutionary algorithm. To facilitate the process of inverse modeling, the objective space is partitioned into several subregions by predefining some reference vectors. In the original work, without loss of generality, the reference vectors are uniformly distributed, while in practice, it is found that for some MOPs which have irregular (nonuniform/disconnected) Pareto fronts (PFs), uniformly distributed reference vectors may lose efficiency. To this end, an adaptive reference vector generation strategy is proposed to deal with the MOPs which have irregular PFs. The basic idea of the proposed strategy is to dynamically adjust the reference vectors according to the distribution of the candidate solutions in the objective space. Specifically, the proposed strategy consists of two phases in the search procedure. In the first phase, the adaptive strategy enhances the population diversity for better exploration; while in the second phase, the proposed strategy enhances the population convergence for better exploitation. To assess the performance of the proposed strategy, empirical simulations are carried out on some typical DTLZ benchmark problems, including DTLZ5 and DTLZ7, which have a degenerate PF and a disconnected PF, respectively. As a result, the proposed adaptive reference vector strategy show promising performance on MOPs which have irregular PFs.},
	publisher = {IEEE},
	booktitle = {The 8th International Conference on Evolutionary Multi-Criterion Optimization}
}