University of Leiden, The Netherlands
Indicator Based Multiobjective Optimization – Geometrical, Heuristic, and Numerical Algorithms
The term indicator-based multiobjective optimization was coined in 2004 by Eckart Zitzler and Simon Künzli and describes algorithms that are guided by a performance indicator for the Pareto front approximation, such as the hypervolume indicator or the Hausdorff metric. While originally the range of such algorithms was confined to metaheuristics, recently the first and second order numerical algorithms and hybridized, memetic algorithms are being developed that are based on similar principles and provide guarantees of local convergence. Moreover, in the last decade many properties of indicators were obtained, such as results on the computational complexity and optimal distribution of points on the Pareto front. The results are often surprising and, besides being interesting for algorithm design, reveal deep insights into the geometry of low dimensional orthoconvex polytopes.
Michael T. M. Emmerich is Associate Professor at the Leiden Institute of Advanced Computer Science. There, he is leader of the Multicriteria Optimization and Decision Analysis (MODA) Group. He is known for co-developing algorithms on indicator based multiobjective optimization, such as the SMS-EMOA, Hypervolume Gradient Ascent and the Hypervolume Newton Method. Moreover he has contributed to the geometrical understanding and to the complexity theory of indicator computations and
developed algorithms on surrogate-assisted black box optimization. He has worked at various institutes in Germany (TU Dortmund, Center for Systems Analysis/ICD e.V., RWTH Aachen), in Portugal (IST Lisbon, Univ. Algarve) and in the Netherlands (FOM/AMOLF, Leiden University) and organized several international workshops and conferences in his research field. He is co-author of more than 140 research publications of which five received best paper awards. Moreover, he was invited speaker at several international events, e.g., Evolutionary Multiobjective Optimization, IEEE Summer School on Multiobjective Optimization, and at renowned universities (Princeton University, Cambridge UK). As a teacher he is involved with teaching classes on optimization theory and on the theory of complex networks.