Seminarium: Tightening upper bounds on components of Pareto optimal outcomes for large-scale multiobjective MIP problems

05/11/2021

Zapraszamy na wspólne seminarium naukowe Pracowni Inteligentnych Systemów Wspomagania Decyzji Politechniki Wrocławskiej, Zakładu Wspomagania Decyzji w Warunkach Ryzyka IBS PAN oraz Centrum Technik Informatycznych WIT.

W dniu 10 listopada (środa) o godzinie 13:15 odbędzie się wspólne seminarium naukowe Pracowni Inteligentnych Systemów Wspomagania Decyzji Politechniki Wrocławskiej, Zakładu Wspomagania Decyzji w Warunkach Ryzyka IBS PAN oraz Centrum Technik Informatycznych WIT, na którym dr inż. Janusz Miroforidis z Zakładu Wspomagania Decyzji w Warunkach Ryzyka IBS PAN wygłosi referat pt. „Tightening upper bounds on components of Pareto optimal outcomes for large-scale multiobjective MIP problems”.

Tightening upper bounds on components of Pareto optimal outcomes for large-scale multiobjective MIP problems

Abstract:

Despite the rapid development of optimization techniques, there are still multiobjective optimization problems for which the derivation of Pareto optimal solutions can be time and memory-consuming or even beyond allotted limits. This applies, for example, to multiobjective MIP problems, where the "curse of dimensionality" hits even top-class commercial solvers. At the previous seminar, we showed a general methodology for multiobjective optimization to provide lower and upper bounds on objective function values of a Pareto optimal solution whenever it cannot be derived within a reasonable time limit on optimization. At this seminar, we show how to provide such bounds of different tightness by problem relaxations, depending on how much time one is willing to spend on extra computations. In any case, we are in a position to keep those extra computations within a small fraction of the time spent on deriving approximations of Pareto optimal solutions. To illustrate the approach viability, we present results of several numerical experiments with large-scale multidimensional 0-1 knapsack problems with two and three objective functions.

Seminarium będzie przeprowadzone w trybie stacjonarnym na Politechnice Wrocławskiej w budynku C-3, ul. Janiszewskiego 11, sala 118. oraz za pomocą systemu telekonferencyjnego.

Link do spotkania

Seminarium: Tightening upper bounds on components of Pareto optimal outcomes for large-scale multiobjective MIP problems