Modelling uncertainty in optimization: illustration on the knapsack problem
28 Novembro 2013 ·
- Quem: Michael Poss (CNRS)
- Onde: FGV - Praia de Botafogo, 190, room 317
- Quando: 28 de Novembro de 2013 às 16:00h
A large variety of decision problems can be accurately modeled as linear optimization problems that involve decisions that can be chosen among discrete sets of possibilities (MILP). Moreover, decision makers are usually facing several source of uncertainty which are translated into uncertain parameters in the optimization models. A simplistic way to handle these uncertainties is to overestimate the unknown parameters. However, this results in wasting capital as well as operational expenditures such as energy resources. In contrast to this naive approach, researchers have introduced frameworks where the uncertainty is directly taken into account either by considering random variables as input (and optimizing expected costs) or in a worst case approach where the uncertainty parameters are constrained in a convex set. These frameworks are respectively denoted by Stochastic Programming and Robust Optimization (RO).
In this talk, we will introduce these methodologies and describe how they can be applied to the knapsack problem, a very important MILP. We will then present some of our recent works and show how to solve efficiently different versions of stochastic and robust knapsack problems. Our results are both theoretical and empirical.
Biography
Michael Poss graduated in Mathematics from the Université Libre de Bruxelles (Belgium) and in Operational Research from the University of Edinburgh (UK). He obtained his PhD in Operations Research from the Université Libre de Bruxelles in February 2011, after which he spent a couple of months at the Universidade de Aveiro (Portugal), followed by a postdoctoral stay at the Universidade de Coimbra (Portugal). He is a CNRS researcher at Heudiasyc since October 2012. His interests are in Operations Research, and more particularly, Optimization under Uncertainty.
Important
- Not required a confirmation.
- FGV not allow the entrance of people wearing shorts and/or slippers.