tag:blogger.com,1999:blog-593563533834706486.post5554480390719700021..comments2024-03-09T06:47:13.003-05:00Comments on Yet Another Math Programming Consultant: Cplex logErwin Kalvelagenhttp://www.blogger.com/profile/09496091402502236997noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-593563533834706486.post-9314045925062008202010-08-10T13:25:53.782-04:002010-08-10T13:25:53.782-04:00As far as i can see from the additional info your ...As far as i can see from the additional info your feeling is spot on and the presolver redeclares the variables investmentdone(i,j,k,t) from continous on [0,1] to binary as they cannot take any values other than 0 and 1.<br />This does give the Solver more information about the model and variables which does make sense - once we know that those variables are in fact binary we dont need to worry about rounding errors for example, branching strategies might be adjusted as well (speculation).<br /><br />You could try to disable presolving (or just MIP presolve, if that is an option?) alltogether and compare performance if you have time to spare.Sebastianhttps://www.blogger.com/profile/09997565096648982268noreply@blogger.comtag:blogger.com,1999:blog-593563533834706486.post-68493634274959593232010-08-01T18:23:02.932-04:002010-08-01T18:23:02.932-04:00This does look interesting indeed. But the size of...This does look interesting indeed. But the size of the problem in terms of rows, columns and nonzeros is reduced after the second aggregator run. Could it be that cplex was able to restrict 50k integer variables to binaries due to their domain?<br />I would be interested in more details about the model you are running. Can you provide for example a summary of variables and their types before and after presolve?Sebastianhttps://www.blogger.com/profile/09997565096648982268noreply@blogger.comtag:blogger.com,1999:blog-593563533834706486.post-34778897797529151022010-08-01T18:16:11.396-04:002010-08-01T18:16:11.396-04:00I sometimes find that with certain formulations, m...I sometimes find that with certain formulations, more binaries doesn't necessarily mean more difficult. A notable example is when using convex hull formulations for disjunctive programming.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-593563533834706486.post-88753356700303046062010-07-31T20:18:57.204-04:002010-07-31T20:18:57.204-04:00Does it solve faster with the aggregator turned of...Does it solve faster with the aggregator turned off than with it on?Paul A. Rubinhttps://www.blogger.com/profile/05801891157261357482noreply@blogger.com