HEC Montréal, Canada, May 2 - 4, 2011

2011 Optimization Days

HEC Montréal, Canada, 2 — 4 May 2011

Schedule Authors My Schedule

TB2 Optimisation continue / Continuous Optimization

May 3, 2011 01:30 PM – 03:10 PM

Location: Béton Grilli

Chaired by Dominique Orban

4 Presentations

  • 01:30 PM - 01:55 PM

    Solving Unconstrained Non-convex Programs Using ACCPM

    • Ahad Dehghani, presenter, GERAD - McGill University
    • Dominique Orban, GERAD - Polytechnique Montréal
    • Jean-Louis Goffin, GERAD - McGill University

    The analytic center cutting plane method (ACCPM) and proximal ACCPM are well known techniques for solving linear and nonlinear convex programming problems. We propose two sequential convex programming methods based on ACCPM and con- vexification techniques to tackle unconstrained problems with a non-convex objective function.

  • 01:55 PM - 02:20 PM

    Fast Local Convergence of Interior-Point Methods in the Absence of Strict Complementarity

    • Zoumana Coulibaly, presenter, GERAD
    • Dominique Orban, GERAD - Polytechnique Montréal

    We show that whenever the strict-complementarity assumption fails to be satisfied at a local solution, an appropriate scaling of the primal-dual Lagrange multiplier estimates allows to recover superlinear convergence in interior-point methods for nonlinear optimization.

  • 02:20 PM - 02:45 PM

    Inexact Optimization for Multidisciplinary Aerospace Design

    • Andrew Lambe, presenter, University of Toronto
    • Joaquim Martins, University of Michigan

    Many optimization problems found in aerospace design contain thousands of design variables and constraints. We present an optimization strategy based on inexact Newton methods for efficiently solving problems of this size. Results from several test problems show the effectiveness of the inexact method compared to an exact strategy.

  • 02:45 PM - 03:10 PM

    Iterative Methods for SQD Linear Systems

    • Dominique Orban, presenter, GERAD - Polytechnique Montréal
    • Mario Arioli, Rutherford Appleton Laboratory

    Symmetric quasi-definite (SQD) systems arise naturally in
    interior-point methods for convex optimization and in some regularized
    PDE problems. In this talk we review the connection between SQD linear
    systems and other related problems and investigate their iterative