Duration: 2 hours
Start: September 05, 16:30

Location: Auditorium "José Adem"


Abstract:
In many applications one is faced with the problem that several conflicting objectives have to be optimized concurrently. One important characteristic of such multi-objective optimization problems (MOPs) is that they typically do not have one single solution as "classical" scalar optimization problems (SOPs). Instead, one can expect that the solution set of a MOP -- the so-called Pareto set -- forms at least locally a manifold of dimension k-1, where k is the number of objectives considered in the MOP. For the treatment of MOPs specialized evolutionary algorithms, called multi-objective evolutionary algorithms (EMOAs), have caught the interest of many researchers and practitioners during the last three decades. Reasons for this include that EMOAs are applicable to a wide range of problems, require minimal assumptions on the model, and are of global nature. Further, due to their set based approach they allow to compute a finite size representation of the entire Pareto set in a single run of the algorithm. Crucial for the success of the EMOAs, however, is the proper archiving. Archiving (also called selection) refers to the strategies that decide which of the computed candidate solutions are kept during the run of the algorithm and which ones are discarded.

In this tutorial, we will first give a general introduction to multi-objective optimization. In the second part, we will address multi-objective archiving. To this end, we will present several such strategies that aim to generate suitable finite size approximations of the Pareto set, respectively its image, the Pareto front, as well as other, related, sets of interest. The methods will be analyzed with respect to their monotonic and limit behavior. The latter will be done using a broad framework. All of the presented archivers can effortlessly be coupled with any set-based multi-objective search algorithm such as MOEAs, and the resulting hybrid method takes over the convergence properties of the chosen archiver. We close the talk with a discussion of possible paths of future research in this direction.

Literature:
O. Schuetze and C. Hernandez
Archiving Strategies for Evolutionary Multi-objective Optimization Algorithms, Springer, 2021
https://link.springer.com/book/10.1007/978-3-030-63773-6

Contact:

Dr. Oliver Schütze This email address is being protected from spambots. You need JavaScript enabled to view it.
https://neo.cinvestav.m/Group

Oliver Schütze received a PhD in Mathematics from the University of Paderborn, Germany, in 2004. He is currently professor at the Cinvestav-IPN in Mexico City, Mexico. The research interests of Dr. Schütze focus on numerical and evolutionary optimization where he addresses both numerical and evolutionary techniques. He has co-authored more than 170 publications including 2 monographic books, 5 text books and 16 edited books. Google Scholar reports more than 4,400 citations and a Hirsch index of 35. During his career he received several prices and awards. For instance, he is co-author of two papers that won the IEEE CIS Outstanding Paper Award (for the IEEE TEC papers of 2010 and 2012). He is recipient of the C. S. Hsu Award 2022. He is Editor-in-Chief of the journal Mathematical and Computational Applications, and member of the Editorial Board for IEEE Transactions on Evolutionary Computation, Applied Soft Computing, Computational Optimization and Applications, Engineering Optimization, and Research in Control and Optimization. Dr. Schuetze is member of the Mexican Academy of Sciences (AMC) and the National Network of Researchers (SNI Level III). More information about Dr. Schütze and his research team can be found at: https://neo.cinvestav.mx/Group/index.php