Charla de Sébastien Verel
Sobre la aplicación de modelos subrogados para problemas de optimización definidos usando gemelos digitales. Como parte de la asignatura "Algoritmos Evolutivos", del Máster en Ingeniería del Software e Inteligencia Artificial
Categoría: I+D+i UMA
PONENTE: Dr. Sébastien Verel, de la Université du Littoral Côte d'Opale (Francia).
FECHA: Martes, 29 de noviembre de 2022, 16:00
LUGAR: Online en el enlace https://eu.bbcollab.com/guest/8a0387ce33f743379e62234cdd07e93c
TÍTULO: "Optimization for simulation based problems"
RESUMEN: The need of frugal products, services with high quality and resilience, or more generally new efficient systems can take the avantage of using digital twin during the design phase. A digital twin which is a numerical model allows to try, and test new potential designs among thousands with a reduced cost, and higher safety. Then, black-box optimization techniques such as local searches or evolutionary algorithms, can be used to find efficient new design by simulating the digit twin. In this case, one of the main difficulty is the computation time of the simulation which can be high. This talk will give two examples of simulation based problems, and different approaches to solve them. In particular, surrogate model approach will be present which learn a model of the simulation in order to accelerate the search throw the optimization search space.
BIOGRAFÍA: Dr. Sébastien Verel is Professor at the Université du Littoral Côte d’Opale, where he is Director of LISIC lab and responsible of the Master 2 in Computer Science. He was also Director of the Computer Science teaching department from 2017 to 2020. Dr. Verel has been co-organizer of 15 workshops/sessions on "Evolutionary Multiobjective Optimization" and "Fitness landscape analysis", co-track chair of the "Evolutionary Combinatorial Optimization and Metaheuristics" track in the Genetic and Evolutionary Computation Conference and Program chair of the EvoCOP conference. Dr. Verel has been co-editor of the special issue on "Evolutionary Multiobjective Optimization" in the European Journal of Operations Research, co-organizer of the Artificial Evolution Summer School in four editions and has offered tutorials in the main conferences on Evolutionary Computation: CEC, PPSN, GECCO and LION. Sébastien Verel is the author of more than 140 publications and has contributed to the Paradiseo software library and the MoCObench benchmark for Multiobjective Combinatorial Optimization.