g²

[g² @geraldgamrath]

@TedRalphs @INFORMS Congratulations, Ted. Very well deserved!

[T-mo B-hold @timoberthold]

. @localsolver

[T-mo B-hold @timoberthold]

Big News at the #SCIPOPT Workshop. As of today, dual licensing is history, and SCIP is fully free software, distributed under the Apache 2.0 license @AchterbergT @ambrosgleixner #orms

[g² @geraldgamrath]

@ambrosgleixner @schmaidt @svigerske @MattMilten In the MIPLIB 2010 paper, we reported that the virtual best (incl. CPLEX and Xpress) was a factor >2 faster than the individual ones. I'd say that current MINLP solvers use more different approaches and are more specialized, which benefits the virtual best, as the factor 5 shows.

[g² @geraldgamrath]

@ambrosgleixner @schmaidt @svigerske @MattMilten I guess the virtual best for MIPLIB suffers from CPLEX, Xpress and SAS OR not being benchmarked by Hans anymore. In the MINLP comparison, there are 3 other solvers within a factor of 2.5 to SCIP; for MIPLIB, only COPT is that close and the next one already has a factor of ~8.

[g² @geraldgamrath]

@NikosKazazakis @AchterbergT @ambrosgleixner @MattMilten @Octeract_UK But surely your results will change if the columns or rows are permuted, right?

[g² @geraldgamrath]

@RobertCSchwarz @timoberthold Yes, you can see this in the line diff-timeouts, which considers the instances where exactly one of the versions timed out.

[g² @geraldgamrath]

50+ people in the audience enjoying the presentation and waiting for news from Gurobi, SAS OR, and COPT.

[g² @geraldgamrath]

"To cut or not to cut?" @timoberthold presenting news about Xpress Optimizer at #euro2022espoo

[g² @geraldgamrath]

@ZuseInstitute @LNDWBerlin 2. Juli

[g² @geraldgamrath]

@MattMilten Too bad, you missed the update of COPT to our new version 1.4 by a few hours. I would love to see the updated plot. 😄

[T-mo B-hold @timoberthold]

CO@Work moves Online! Our 2-week summer school on Mathematical Optimization will be organized via YouTube lectures and Zoom exercises, in a way that covers all time zones worldwide. Learn more (and register) at https://t.co/5udWDuYGrv
👩‍💻🧑‍💻👨‍💻👩‍💼👨‍🏫➡️👩‍🎓🧑‍🎓👨‍🎓

Please Retweet :-)

[g² @geraldgamrath]

@JeffLinderoth Thank you, @JeffLinderoth! We are very excited to finally release the SCIP Optimization Suite 7.0. It will be online later today. It's actually the third time in a row that Optimization Online had to increase the max number of authors for a SCIP release report. 😀 #orms

[g² @geraldgamrath]

Very happy to have had 50+ spectators, wouldn't have been a real defense without you! Hope you enjoyed it, I definitely did! Thank you to my great referees @thor_koch and @dominiqs81!

[g² @geraldgamrath]

Less than 12 hours left until my PhD defense starts (10 am). Mostly online, thanks to #COVID19: https://t.co/kdx1OqLIpC

[g² @geraldgamrath]

@mluebbecke @mgalati13 @localsolver @opensourcesblog @bbrunaud @JDarlay Ah, sorry, we divide by the minimum absolute value, so for negative objectives values, it's the primal bound in case of minimization.

[g² @geraldgamrath]

@mgalati13 @mluebbecke @localsolver @opensourcesblog @bbrunaud @JDarlay The problem with gap is that there are different definitions around (divide by primal bound/dual bound/minimum of the two). The only thing that everyone treats the same is gap 0, that's why we went for that.

[g² @geraldgamrath]

@mluebbecke Wow, 8 in the morning. Good that I was (still?) awake enough to choose a useful commit message. 😁
Very proud and honored that the code I started for my Diploma thesis is still so actively developed at @OR_RWTH

[g² @geraldgamrath]

@natebrix @mluebbecke @JeffLinderoth How about checking out Hans' results on MIPLIB 2010: https://t.co/GLjqajkJTJ
MIPLIB 2017 has only been released today, after all, perhaps people can wait a few more days until they can look up final results online. #KnightsOfMIP #fairbenchmarking