tag:blogger.com,1999:blog-8781383461061929571.post5608852419371594924..comments2024-03-14T09:08:19.035-04:00Comments on OR in an OB World: Semicontinuous VariablesPaul A. Rubinhttp://www.blogger.com/profile/05801891157261357482noreply@blogger.comBlogger11125tag:blogger.com,1999:blog-8781383461061929571.post-4228585700661433382019-11-30T17:15:43.004-05:002019-11-30T17:15:43.004-05:00Per an inside source, if CPLEX deems the bound on ...Per an inside source, if CPLEX deems the bound on a semicontinuous variable to be reasonable, it linearizes using a binary variable (effectively creating a big-M constraint). If the bound is deemed too large, it uses an indicator constraint. How the indicator constraint is then handled can be seen in the quote from Ed Klotz embedded in an answer by Mark L. Stone to a question about indicators in CPLEX: https://or.stackexchange.com/questions/231/when-to-use-indicator-constraints-versus-big-m-approaches-in-solving-mixed-int/348#348.Paul A. Rubinhttps://www.blogger.com/profile/05801891157261357482noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-52648326249330506172019-11-28T11:41:04.294-05:002019-11-28T11:41:04.294-05:00Has anybody found out for sure how cplex treats se...Has anybody found out for sure how cplex treats semicontinuous variables? Is it using auxillary binary variables or not?Anonymoushttps://www.blogger.com/profile/10495458753837205156noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-20449566847542823282011-04-06T16:11:28.708-04:002011-04-06T16:11:28.708-04:00@Jensen: I am more concerned with job search becas...@Jensen: I am more concerned with job search becasue, I faced lot of difficulties in getting an OR job. I still have some of my friends searching for a job in OR. It was very difficult for me to make an unknown person in linkedin to forward my resume. But, now I am not hesitating to forward someones resume. What they need is just an interview call through which they can prove themselves.bharathhttps://www.blogger.com/profile/01735024811188875205noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-38960374498437948682011-04-01T08:13:21.277-04:002011-04-01T08:13:21.277-04:00@bharat I understand your request for OR job listi...@bharat I understand your request for OR job listing, sadly there are few (or none) really good listings. The information is squattered in various places and hard to find since the traditional key words is not really useful. Often companies look for other skills combined with OR.Bo Jensenhttps://www.blogger.com/profile/10533779186894174073noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-17271682797834553142011-03-25T17:29:31.626-04:002011-03-25T17:29:31.626-04:00@bharath: Thanks for the kind words. As it turns ...@bharath: Thanks for the kind words. As it turns out, I'm among the last people on the planet to hear about OR job openings -- we don't offer any OR degrees here (we did 20+ years ago), so we don't have any recruiting for OR positions here. INFORMS has a "jobs bank" (http://www.informs.org/Build-Your-Career/OR-MS-Job-Bank), and there are occasional postings on LinkedIn either tagged OR or posted in the INFORMS group there. There was some discussion of job listings on OR-Exchange a while back, too.Paul A. Rubinhttps://www.blogger.com/profile/05801891157261357482noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-15037032838679401812011-03-25T17:00:25.630-04:002011-03-25T17:00:25.630-04:00Paul,
I came across your blog just today. It is a...Paul,<br /><br />I came across your blog just today. It is a great source for OR people. I am working as a OR specialist for a marketing company. What I would also like to see in your blog is a section for OR jobs that u come across through your friends. It would help lot of people searching for OR jobs.bharathhttps://www.blogger.com/profile/01735024811188875205noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-53969910067861365672011-03-23T18:10:36.596-04:002011-03-23T18:10:36.596-04:00The LP relaxation apparently just treats the lower...The LP relaxation apparently just treats the lower bound as zero. At least, if I set up a model with a semicontinuous variable and then add a conversion object to convert it to an ordinary float variable, the upper bound is preserved and the lower bound is set to zero.Paul A. Rubinhttps://www.blogger.com/profile/05801891157261357482noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-31496260961165681052011-03-22T19:57:25.288-04:002011-03-22T19:57:25.288-04:00Yeah, that's what i thought. But i must confes...Yeah, that's what i thought. But i must confess that i used this feature practically 2-3 yrs ago, and it must have been least 2 versions prior to 12.x (we didn't have access to latest at that time). One would expect that by now, they would have implemented the best proven approach that can be practically integrated into an already complex code.<br /><br />Perhaps printing out the CPLEX-constructed .lp relaxation for a small example may give us a clue.Shivahttps://www.blogger.com/profile/05571015480979394896noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-49319728712182508172011-03-21T19:59:01.978-04:002011-03-21T19:59:01.978-04:00@Shiva: Is it your impression that CPLEX inserts t...@Shiva: Is it your impression that CPLEX inserts the auxiliary binary variable (and constraints) when it sees a semicontinuous variable? I thought it was branching on the condition (your third point). If so, that raises the question of how the sc variable's domain factors into LP relaxations prior to the branch (is it treated as having domain [0, b] until then, which would be my guess).Paul A. Rubinhttps://www.blogger.com/profile/05801891157261357482noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-26057786126888097282011-03-21T19:53:27.654-04:002011-03-21T19:53:27.654-04:00sorry, pt3 should read "... rather than on th...sorry, pt3 should read "... rather than on the auxiliary binary variables y".Shivahttps://www.blogger.com/profile/05571015480979394896noreply@blogger.comtag:blogger.com,1999:blog-8781383461061929571.post-12949710928134144252011-03-21T19:52:19.581-04:002011-03-21T19:52:19.581-04:00My 2 cents to add from past experience with the se...My 2 cents to add from past experience with the semi-cont approach in CPLEX on some practical examples. <br /><br />1. It is a cool and useful feature to have, but should be used with caution due to the 'big M' like structure hidden inside the resultant MIP. <br /><br />2. In many practical instances, the naive LP relaxation is weak. <br /><br />3. So, if possible, better to branch directly on the semi-cont. conditions (Nemhauser and Farias) rather than on the auxiliary binary variables x.Shivahttps://www.blogger.com/profile/05571015480979394896noreply@blogger.com