2018 Optimization Days

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

Schedule Authors My Schedule
Cal add eabad1550a3cf3ed9646c36511a21a854fcb401e3247c61aefa77286b00fe402

WAP N. Mladenović

May 9, 2018 09:00 AM – 10:00 AM

Location: Plenaries - Amphithéâtre Banque Nationale

Chaired by Gilles Caporossi

1 Presentation

  • Cal add eabad1550a3cf3ed9646c36511a21a854fcb401e3247c61aefa77286b00fe402
    09:00 AM - 10:00 AM

    Less is more approach in optimization

    • Nenad Mladenovic, presenter, Mathematical Institute SANU
    • Daniel Aloise, Polytechnique Montréal
    • Jack Brimberg, GERAD, The Royal Military College of Canada
    • Dragan Urosevic, Mathematical Institute SANU

    Whichever creative work a man undertakes, if the inclination to improve it by adding more and more new elements prevails, there comes a moment when he/she reveals that the obtained result is far from the desired and expected. It is the case in almost all the scientific and art disciplines, in Architecture, Music, Physics, Medicine, Neurosciences, Teaching, Cuisine, etc. A respond to this "more and more" attitude is an approach usually called "Less is more". My collaborators and I have recently proposed "Less is more approach" (LIMA) in Optimization. Its main idea is to find the minimum number of search elements (ingredients) when solving an optimization problem that would make some optimization method more efficient and effective than the currently best. LIMA has appeared as a reaction to more and more complex hybrid heuristic methods that combine many different ideas yet giving no proper explanations for such combinations. Combining several heuristics to get a new hybrid method has a price of losing efficiency and user friendliness, the two very important and desired properties of any heuristic. Indeed, despite of the simplicity of LIMA, we got significantly better results than the more complex heuristics have got in solving several classical optimization problems. Such examples will be presented in my talk. Thus, including many ideas in the search does not necessarily lead to better computational results. On the contrary, sometimes less can yield more.