BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Axel Parmentier (École des Ponts ParisTech) DTSTART:20200616T180000Z DTEND:20200616T183000Z DTSTAMP:20260423T021901Z UID:DOTs/13 DESCRIPTION:Title: Le arning to approximate industrial problems by operations research classic p roblems\nby Axel Parmentier (École des Ponts ParisTech) as part of Di screte Optimization Talks\n\n\nAbstract\nPractitioners of operations resea rch often consider difficult variants of well-known optimization problems\ , and struggle to find a good algorithm for their variants while decades o f research have produced highly efficient algorithms for the well-known pr oblems. We introduce a "machine learning for operations research" paradigm to build efficient heuristics for such variants of well-known problems. I f we call the difficult problem of interest the hard problem\, and the wel l known one the easy problem\, we can describe our paradigm as follows. Fi rst\, use a machine learning predictor to turn an instance of the hard pro blem into an instance of the easy one\, then solve the instance of the eas y problem\, and finally retrieve a solution of the hard problem from the s olution of the easy one. Using this paradigm requires to learn the predict or that transforms an instance of the hard problem into an instance of the easy one. We introduce two structured learning approaches to learn this p redictor\, and illustrate our paradigm and learning methodologies on sever al scheduling and path problems.\n LOCATION:https://researchseminars.org/talk/DOTs/13/ END:VEVENT END:VCALENDAR