TU Delft teams get 1st and 2nd place at the GECCO 2020 Industrial Challenge
A team of students participating in the course Algorithms for Intelligent Decision Making, namely Samuel Rosas, Nick Cleintuar, Rebecca Gedda, and Rickard Karlsson, have won the 1st place in the GECCO 2020 Industrial Challenge (limited budget track). Former Computer Science student Arthur Guijt has won 2nd place. Both teams were supervised by Laurens Bliek, Sicco Verwer, and Mathijs de Weerdt.
The challenge consisted of solving an expensive optimization problem, where each evaluation of the objective function consisted of running a computationally intensive fluid dynamics simulation. The goal was to design part of an industrial gas cleaning system in order to reduce environmental pollution as much as possible, using only a limited budget of function evaluations. Besides the expensive objective, a big challenge was the size of the search space, where over 10^44 design choices were possible. Because of the expensive objective function, both teams used a surrogate-based optimization algorithm (called IDONE) developed by the supervisors that optimizes an efficient approximation of the objective. Both teams made a different adaptation to this algorithm so that it better explored the search space.
More information on the challenge can be found at https://www.th-koeln.de/informatik-und-ingenieurwissenschaften/gecco-challenge-2020_72989.php.