On the Use of Dynamic Reference Points in HypE
In evolutionary multiobjective optimization, hypervolume indicator is one of the most commonly-used performance metrics. To reduce its high computational costs in many objective optimization, Monte Carlo method is used in HypE (Hypervolume Estimation algorithm for multi-objective optimization) for approximating hypervolume values. However, the diversity preservation of HypE can be poor under inappropriate settings of the reference point. In this paper, the influence of the reference point on HypE is discussed and two variants of HypE algorithm with dynamic reference points are proposed to improve the performance of HypE. Our experimental results suggest that the new algorithms outperform HypE with fixed reference points on a set of multiobjective test instances with different shapes of Pareto fronts.
KeywordsMultiobjective optimization Evolutionary computation Hypervolume Reference point
The work described in this paper was supported by the National Science Foundation of China under Grant 61473241, and a grant from ANR/RCC Joint Research Scheme sponsored by the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. A-CityU101/16) and France National Research Agency (ANR-16-CE23-0013-01).
- 1.Auger, A., Bader, J., Brockhoff, D., Zitzler, E.: Theory of the hypervolume indicator: optimal \(\mu \)-distributions and the choice of the reference point. In: FOGA 2009: Proceedings of the Tenth ACM SIGEVO Workshop on Foundations of Genetic Algorithms, pp. 87–102 (2009)Google Scholar
- 4.Bringmann, K., Friedrich, T.: Approximating the volume of unions and intersections of high-dimensional geometric objects. In: Hong, S.-H., Nagamochi, H., Fukunaga, T. (eds.) ISAAC 2008. LNCS, vol. 5369, pp. 436–447. Springer, Heidelberg (2008). doi: 10.1007/978-3-540-92182-0_40 CrossRefGoogle Scholar
- 9.Ishibuchi, H., Imada, R., Setoguchi, Y., Nojima, Y.: Hypervolume subset selection for triangular and inverted triangular pareto fronts of three-objective problems. In: Proceedings of the 14th ACM/SIGEVO Workshop on Foundations of Genetic Algorithms, pp. 95–110 (2017)Google Scholar
- 10.Ishibuchi, H., Tsukamoto, N., Sakane, Y., Nojima, Y.: Hypervolume approximation using achievement scalarizing functions for evolutionary many-objective optimization. In: 2009 IEEE Congress on Evolutionary Computation, pp. 530–537. IEEE (2009)Google Scholar