Richard Oberdieck

Finally a new blog post! This one is about #python and a pattern I use on the daily to deal with heavy inits. Disagree? Leave me a comment! oberdieck.dk/p/create-patte…

@e_d_andersen Interesting. Is it about him specifically, or generally the whole atomic bomb stuff? I've been very interested in von Neumann for the past few months, so this might be an interesting addition.

Interested in John Von Neumann then I can reconmend

@themarklstone Ah, now I understand. I think that is more a reflection of the style of writing that was common in 1940, I didn't find any derogatory comments in the text.

@ROberdieck My question was inspired by quotes in your review: "which may lead men to prosecute research" and "“A Mathematician ... and so his patterns" and "applied mathematician ... If he wants to be useful, he must work in a humdrum way, and he cannot give full play to his fancy ..."

Looking for a book for the holidays and you are into maths? Then check out the famous essay by Hardy, "A Mathematician's Apology". Really a fun one to read if you like the quirkiness of mathematicians :) My notes in my book club: oberdieck.dk/book-club/hard…

@e_d_andersen Hm, never paid attention but this should be easy to test. I just answered what I got using Gurobi (gotta try with MOSEK as well ;-) ). I treat MPS files as black boxes most of the time. The only person I have ever seen read MPS files is @Ed_Klotz, the rest of us read LP files :D

I am confused about the specfication of the integer constrained variables in the MPS format. or.stackexchange.com/questions/1126…

@themarklstone It's almost as if he didn't think of people as valuable, but only of the things they produce. And the more permanent and influential they are, the better. Gender, race, age etc. does not matter for as long as the result is eternal. [that's just my opinion] 2/2

@themarklstone I don't remember him singling out female mathematicians specifically. I'll have another look and let you know. A quick google search didn't bring anything up. If I had to guess, I don't think he would've cared for as long as the results created are beautiful and unique. 1/2

I also started a book club, where I will share my notes of books I am reading. The first one is on hacking by @scottjshapiro, find it at oberdieck.dk/book-club/fanc…

Finally got around to put a new post together, this time about "beautifulsoup4", a Python library for web scraping. oberdieck.dk/p/web-scraping… It really is awesome what one can accomplish with Python in a short amount of time! #python #beer #scraping

@dirk_sch Uhh, can I use that one in my presentations (with reference to you)? Love it!

Everyone working on O.R. problems

@EliasHasle Btw, regarding the "trash" part: I grudgingly admit that metaheuristics are sometimes(!!) the right choice. When a problem is very complex, doesn't change a lot and would be hard to model otherwise (e.g. due to a blackbox model), I might throw a GA or ACO at it to see how it goes

@ROberdieck The algorithm is written by humans, though. 🙄

Just stumbled upon this list of metaheuristics again. The naming is really unbelievably imaginative...have fun! github.com/fcampelo/EC-Be…

@EliasHasle 😂gotta love the fact that they call it a "human-based" approach, when it, in fact, uses a computer...

@ROberdieck It seems to me that most metaheuristics papers are trash. It's like the name is the game. x.com/EliasHasle/sta…

@AustinLBuchanan @HighsOpt @pypi I just checked and the API is not super-easy. For example, there is no "addConstr", but an "addRow". @FICO_Xpress has the same and it makes it quite quick but it's not really nicely readable. Would be cool if someone does a port to python-mip. See here ergo-code.github.io/HiGHS/stable/i…

@HighsOpt is now on @pypi with "pip install highspy", and you got the best open-source solver at your fingertips. I think this is great news for everyone who doesn't have a commercial solver available and still wants to do #optimization! #opensource #highs #ai

@GregorJ4 @HighsOpt @pypi Totally agreed, but if one does not have a commercial solver available, then it's good to have. Been using python-mip with CBC for those cases so far, will give HiGHS a go for my next project.

@ROberdieck @HighsOpt @pypi Highs solver can be best free mip/lp solver but for many problems especially in discrete domain exist other free solvers that can compete with commercial solvers. For example cp-sat from or-tools.

See github.com/ERGO-Code/HiGHS for the details!

Happy to share that I just joined @mosektw as an Advisory Board Member! Looking forward to working with the other board members and @e_d_andersen!

@nandofioretto Uh I'm looking forward to reading this one!

📢 Happy to share our survey on integrating prediction and optimization in end-to-end differentiable systems! 🚀 We dive into foundations, propose a taxonomy, create an extensive benchmark, and share some opportunities in this exciting field! Paper link: arxiv.org/abs/2307.13565

@JuMPjl has responded to the @GamsSoftware blog post from 2 weeks ago: jump.dev/2023/07/20/gam…. They get a 30x speed-up by reformulating the code, so check it out! Cool stuff! The corresponding PR is also an interesting read: github.com/jump-dev/jump-…

Link to source: gurobi.com/case_studies/c… and p1.aprimocdn.net/gurobi/ca9b3b9…

Careful @gurobi, this title is a bit misleading: "Copenhagen Airport optimized with Gurobi, saving $70M USD", while the text then says "which served as the basis for a set of new investments in stands and gates whose total cost was over $70M USD (400M DDK)." Savings != Cost ;-)