Journées de l'optimisation 2018

HEC Montréal, Québec, Canada, 7 — 9 mai 2018

Horaire Auteurs Mon horaire

TB5 Stochastic, robust, and noisy optimization

8 mai 2018 15h30 – 17h10

Salle: Manuvie (54)

Présidée par Jordan Ninin

4 présentations

  • 15h30 - 15h55

    Parameters estimation of 3D model for viscoelastic polymers: Bayesian approach

    • Lingyu Yue, prés., Laboratory for multiscale mechanics
    • Martin Lévesque, Polytechnique Montréal
    • Jonathan Jalbert, Ecole Polytechnique Montreal
    • Marie-Claude Heuzey, Ecole Polytechnique Montreal

    The parameters identification for viscoelastic materials is always open question as the result of ill-posedness. A Bayes' theorem based statistical method is introduced to select appropriate viscoelastic model and estimate parameters from mechanical test data. The experimental errors are also quantified. The results will be validated by a complex loading test on same sample.

  • 15h55 - 16h20

    Coupling decomposition algorithm with dynamic programming for a stochastic spatial model for medium-term energy management problem

    • Luc Marchand, prés., Université de Sherbrooke
    • Philippe Mahey, Institut Supérieur d'Informatique, Modélisation et de leurs Applications
    • Jean-Pierre Dussault, Université de Sherbrooke

    This talk is on energy management from hydro and thermal sources on a medium time scale with multiple zones with stochastic demand and water inflow. We will discuss the use of decomposition to split the problem spatially, then use dynamic programming to solve the zonal problems. We will compare the results and computation times with the direct solving and will discuss about the complexity growth of the two methods.

  • 16h20 - 16h45

    Solving an inverse integer optimization problem with noisy data using a cutting plane algorithm

    • Mahsa Moghaddass, prés.,
    • Daria Terekhov, Concordia University

    This study develops a method for solving an inverse integer optimization problem when solutions are noisy. We propose a cutting plane algorithm to obtain a cost vector for the forward problem such that the given solution becomes optimal or approximately optimal with the minimum optimality gap.

  • 16h45 - 17h10

    Global optimization with quantified constraints and integral objective function: An interval branch and bound approach

    • Jordan Ninin, prés., GERAD
    • Dominique Monnet, GERAD and Polytechnique Montréal
    • Benoit Clément, ENSTA-Bretagne

    The main benefits of Interval-Analysis-based methods are to cope with non-convexity and heterogenous optimization problems. We illustrate these approachs by solving a structured robust control problems with H_2 and H_infinity constraint and model uncertainties. These problems can be formulated as an optimization problem with integral objective function and quantified constraints.