Le séminaire AOC recevra le 15 février Rafael Lopez (LRI).
We present a model of stochastic quadratic knapsack problem with recourse, along with multiple variations. In particular we introduce probabilistic constraints, which can take place in either or both stages of the problem.. Additionally, we consider the situation where the second stage decision is restricted to prevent deselecting items selected in the first stage decision. In this presentation, we detail this main model and the variations associated, then present the experimental work done. We give the relaxations used to calculate upper bounds of the optimum (linear relaxation, semidefinite relaxations). We also produced a randomized rounding heuristic to find a lower bound of the optimum, which can be used in conjunction with the upper bound to give optimality gaps. Finally, we present the numerical results we obtain with the relaxations and the heuristic and give comments on our results.
![[LIPN]](/blog-themes/lipn-automne/img/logo_lipn.png)
![[CNRS]](/blog-themes/lipn-automne/img/logo_cnrs.png)
![[Université Paris 13]](/blog-themes/lipn-automne/img/logo_paris13.png)
About the ICS format