iiinova — where your digital self keeps evolving

Rolf van der Hulst @rolfvanderhulst.bsky.social · 07/10/2026

I find it problematic that inference costs and costs of some of the human operators for the OpenAI math repository are not reported. Now I'm left wondering whether an equivalent financial investment in human mathematicians would have produced similar results.

020

Reposted by Rolf van der Hulst

Stefan Neumann @neumannstefan.com · 07/10/2026

So, the OpenAI results include: - A proof the unique games conjecture - L = RL - Matrix multiplication in n^(9/4) time - Almost-linear-time exact matching - Integer multiplication faster than n log n - Quasipolynomial algorithms for mean-payoff and parity games Overview: github.com/openai/math/...

github.com

math/overview.pdf at main · openai/math

Contribute to openai/math development by creating an account on GitHub.

1197

Reposted by Rolf van der Hulst

Alexander Kasprzyk @amkasprzyk.bsky.social · 08/09/2026

“Why would you ruin your career? … If you don’t want me to be nice, then I don’t have to be nice.” Thuggish behaviour described on the part of OpenAI. As an academic, do you really want to be in the room with these people? Does what’s described here reflect your standards? #LLM #MathSky

2276

Reposted by Rolf van der Hulst

Ragnar {Groot Koerkamp} @curiouscoding.nl · 28/08/2026

Like seriously, if you use LLMs to write text, you're intentionally DDoS-ing those of us who don't....

252

Rolf van der Hulst @rolfvanderhulst.bsky.social · 12/08/2026

God forbid if you're a woman too, my wife has the same issue and the amount of people and forms which assume that the first last name is mine and (automatically) hyphenate the two words is rather infuriating.

010

Rolf van der Hulst @rolfvanderhulst.bsky.social · 22/07/2026

Fully agree with everything you wrote. I wrote my comment using my experience with the denser graphs from my branch-and-price based solver at github.com/rolfvdhulst/pcog

github.com

GitHub - rolfvdhulst/pcog: Parallelized Coloring of Grahs

Parallelized Coloring of Grahs. Contribute to rolfvdhulst/pcog development by creating an account on GitHub.

030

Rolf van der Hulst @rolfvanderhulst.bsky.social · 22/07/2026

Ah okay. Maybe they are referring to some specific result which is about the quality of solutions or heuristics or so, rather than exact approaches? Or perhaps they were referring to color refinement algorithms for graph isomorphism, which I think is a better example.

130

Rolf van der Hulst @rolfvanderhulst.bsky.social · 22/07/2026

A bit of a sidetrack, but I'm wondering why you claim that graph coloring is easier to solve than predicted by theory: As far as I know state-of-the-art approaches for this problem still cannot prove optimal colorings for most instances with more than a couple of hundred vertices.

130

Reposted by Rolf van der Hulst

eevee @eev.ee · 11/06/2026

and hey it turns out that the industry is quite happy to pay millions of dollars for open source code! just not to its developers. no, all that money goes to fucking GOOGLE AND MICROSOFT because they're the ones operating the vacuum cleaners. fucking ludicrous state of affairs.

317336

Rolf van der Hulst @rolfvanderhulst.bsky.social · 01/05/2026

You realize that the MPS format is the way it is because it was used with punchcards in the 1950's? Don't ask me why a better standard has not appeared, but this at least may give you some context.

010

Rolf van der Hulst @rolfvanderhulst.bsky.social · 15/04/2026

Just saw I skipped over the f(0)=1 condition, so disregard that.

000

Rolf van der Hulst @rolfvanderhulst.bsky.social · 15/04/2026

Surely not, since f(x) = 0 also satisfies f'(x) = f(x).

100

Reposted by Rolf van der Hulst

상일 (Sang-il) @sioum.bsky.social · 27/03/2026

4색 정리 새로운 증명이 arXiv에 올라왔습니다. New proof of the four color theorem by Yuta Inoue, Ken-ichi Kawarabayashi, Atsuyuki Miyashita, Bojan Mohar, Carsten Thomassen, Mikkel Thorup arxiv.org/abs/2603.24880

arxiv.org

The Four Color Theorem with Linearly Many Reducible Configurations and Near-Linear Time Coloring

We give a near-linear time 4-coloring algorithm for planar graphs, improving on the previous quadratic time algorithm by Robertson et al. from 1996. Such an algorithm cannot be achieved by the known p...

0128

Rolf van der Hulst @rolfvanderhulst.bsky.social · 20/03/2026

Just learned that arXiv is becoming independent of Cornell. Statement by arXiv: tech.cornell.edu/arxiv/

tech.cornell.edu

Cornell Tech - arXiv

011

Rolf van der Hulst @rolfvanderhulst.bsky.social · 16/03/2026

I also show that the folding method can be applied to the continuous columns of mixed-integer linear programs, and in some cases can also be used to aggregate integer variables by detecting affine totally unimodular decompsitions, which describe a TU-like structure in the constraint matrix.

000

Rolf van der Hulst @rolfvanderhulst.bsky.social · 16/03/2026

The new methods that I formulate extend the ``LP Folding'' reduction that aggregates permutation symmetries in linear programs to also aggregate reflection symmetries.

101

Rolf van der Hulst @rolfvanderhulst.bsky.social · 16/03/2026

I am happy to share my latest work ``Folding Mixed-Integer Linear Programs and Reflection Symmetries'', where I formulate two new presolving techniques for handling symmetries in linear programs and mixed-integer linear programs. Check out the preprint here: lnkd.in/efgYeVSJ lnkd.in/eGpim7U9

lnkd.in

This link will take you to a page that’s not on LinkedIn

161

Reposted by Rolf van der Hulst

ISMP 2027 – Amsterdam @ismp2027.bsky.social · 27/02/2026

📢 #ISMP2027 comes to Amsterdam! The 26th International Symposium on Mathematical Programming will be held July 25–30, 2027. Join researchers from around the world to discuss advances, challenges, and opportunities in the theory and practice of mathematical optimization. 🔗 ismp2027.mathopt.nl #MOS

ismp2027.mathopt.nl

The 26th International Symposium on Mathematical Programming will be held in Amsterdam, The Netherlands, from July 25th through July 30th, 2027. The event is the symposium of the Mathematical Optimization Society (MOS).

1258

Rolf van der Hulst @rolfvanderhulst.bsky.social · 18/02/2026

I'm not sure about Gurobi, but for SCIP I suspect the natural dates formulation of the single machine scheduling problem with cuts based on release times could work: pure.tue.nl/ws/portalfil...

pure.tue.nl

020

Rolf van der Hulst @rolfvanderhulst.bsky.social · 10/02/2026

Just access of any random memory location. This blog series explains it quite well: www.ilikebigbits.com/2014_04_21_m...

ilikebigbits.com

The Myth of RAM, part I

020

Rolf van der Hulst @rolfvanderhulst.bsky.social · 09/02/2026

I think the standard model of computation is partly to blame here. Especially O(1) memory access is not really realistic in practice and is closer to to O(sqrt(N)) in practice.

120

Reposted by Rolf van der Hulst

ArXiv math.OC Optimization and Control @optb0t.bsky.social · 09/02/2026

📚 New Arxiv Paper Title: Circuit Diameter of Polyhedra is Strongly Polynomial Authors: Bento Natura Read more: arxiv.org/abs/2602.06958

063

Rolf van der Hulst @rolfvanderhulst.bsky.social · 20/01/2026

Of course, most papers do define the variables and their meaning in the text above the model. But I think the latter ordering would still be easier to read if you just want to glance at the model.

000

Rolf van der Hulst @rolfvanderhulst.bsky.social · 20/01/2026

The order 'objective - constraints - variables' that academic papers use to describe mathematical optimization problems is inconvenient, as it places the objective before you even know what the variables are. Wouldn't 'variables - constraints - objective' be a more natural progression?

210

Rolf van der Hulst @rolfvanderhulst.bsky.social · 09/12/2025

Given both your recent results, are you still motivated to investigate the simplex method further, or do you want to move on to new questions and topics?

120

Reposted by Rolf van der Hulst

ArXiv math.OC Optimization and Control @optb0t.bsky.social · 19/11/2025

📚 New Arxiv Paper Title: Geometry of the Reformulation-Linearization-Technique: Domination of Disjunctions Authors: Hugo A. Hof, Matthias Walter Read more: arxiv.org/abs/2511.13805

021

Rolf van der Hulst @rolfvanderhulst.bsky.social · 15/11/2025

I shouldn't be surprised by this, but I just learned that Nvidia and AMD both disable FP64 performance units in their consumer cards so scientists and companies have to buy the high end ones at 5-10x the price. It's a valid business model but unfortunate for hobbyists like me who just want to try.

010

Reposted by Rolf van der Hulst

ArXiv math.OC Optimization and Control @optb0t.bsky.social · 12/11/2025

📚 New Arxiv Paper Title: Linear Programming Hierarchies Collapse under Symmetry Authors: Yuri Faenza, V\'ictor Verdugo, Jos\'e Verschae, Mat\'ias Villagra Read more: arxiv.org/abs/2511.07766

021

Rolf van der Hulst @rolfvanderhulst.bsky.social · 08/11/2025

Mostly deterministic RO with uncertainty sets (rather than distributionally RO), which can be formulated as some kind of Mixed Integer Program (does not need to be linear). Most interesting would be instances derived from real-world applications.

100

Rolf van der Hulst @rolfvanderhulst.bsky.social · 08/11/2025

The only thing I could find is robust-optimization.com/instances/ , which collects instances from a few papers. Unfortunately, there seems to be no standardized data format for robust optimization problems.

robust-optimization.com

Instances

100

Rolf van der Hulst @rolfvanderhulst.bsky.social · 06/11/2025

Does there exist a commonly used public instance benchmark set for robust optimization? Or sets of instances which are frequently used within the literature?

153

Reposted by Rolf van der Hulst

Martin Schmidt @schmaidt.bsky.social · 22/10/2025

This was hard work but now we are happy! The #BOBILib paper has today been accepted for publication in Mathematical Programming Computation! Here's the final version of the preprint: optimization-online.org/wp-content/u...! And here's the website bobilib.org! Spread the word and use it! :-)

0143

Reposted by Rolf van der Hulst

ArXiv math.OC Optimization and Control @optb0t.bsky.social · 15/07/2025

🔄 Updated Arxiv Paper Title: Implied Integrality in Mixed-Integer Optimization Authors: Rolf van der Hulst, Matthias Walter Read more: arxiv.org/abs/2504.07209

052

Rolf van der Hulst @rolfvanderhulst.bsky.social · 09/06/2025

At least on Linux there seem to exist specialized fast timing functions: see stackoverflow.com/a/47003703 , which recommends to use 'clock_gettime()' and claims it only takes 3-5 nanoseconds on their system. I think it is likely that alternative calls exist in other operating systems.

100

Rolf van der Hulst @rolfvanderhulst.bsky.social · 09/06/2025

What kind of crazy FPS are you running at where a syscall is too expensive? In my experience, most syscalls take on the order of 100 - 1000 cycles, which would take less than a microsecond on a modern cpu.

100

Reposted by Rolf van der Hulst

Mathieu Besançon @matbesancon.xyz · 04/06/2025

This is now out, we're super proud to have won the Land-Doig MIP competition prize for our algorithm for mixed-integer quadratic optimization!

1205

Reposted by Rolf van der Hulst

passagemath @passagemath.org · 27/04/2025

passagemath 10.5.27 contains (a preliminary version of) the new package passagemath-cmr, an interface to the Combinatorial Matrix Recognition library by Matthias Walter et al., implementing Seymour's decomposition of TU matrices etc. pypi.org/project/pass... #Python #SageMath #OpenSource #MathSky

pypi.org

passagemath-cmr

passagemath: Combinatorial matrix recognition

032

Rolf van der Hulst @rolfvanderhulst.bsky.social · 06/03/2025

If you are checking out Cargo, I would also check out NPM (node package manager), which is one of the most popular javascript package managers.

110

Rolf van der Hulst @rolfvanderhulst.bsky.social · 21/02/2025

No of course not! I would say the opposite; the fact that it is open source helps because you can actually understand what is happening, as long as you take the time to figure it out.

000

Rolf van der Hulst @rolfvanderhulst.bsky.social · 21/02/2025

I was mostly just talking about understanding the code, not just the parameters. I agree that it is annoying that a lot of parameters are repeated.

100

Rolf van der Hulst @rolfvanderhulst.bsky.social · 21/02/2025

I spent a lot of time figuring out that a '>' should be a '<'... What prompted me to write the comment, however, is the amount of details that are involved. I doubt many users could successfully tune any solver without strong knowledge of its internals.

100

Rolf van der Hulst @rolfvanderhulst.bsky.social · 20/02/2025

After attempting to tune MIR cuts in SCIP for 3.5 days, I understand better why people are willing to pay so much for a commercial MIP solver license.

170

Reposted by Rolf van der Hulst

ArXiv math.OC Optimization and Control @optb0t.bsky.social · 05/02/2025

🔄 Updated Arxiv Paper Title: Sparsity-driven Aggregation of Mixed Integer Programs Authors: Liding Xu, Gioni Mexi, Ksenia Bestuzheva Read more: arxiv.org/abs/2502.01192

011

Reposted by Rolf van der Hulst

Marco Lübbecke @mluebbecke.bsky.social · 03/02/2025

registration for the 2025 school on column generation is now open www.gerad.ca/colloques/Co... #orms

gerad.ca

2025 School on Column Generation

0118

Reposted by Rolf van der Hulst

Marco Lübbecke @mluebbecke.bsky.social · 31/01/2025

The bugfix release 9.2.1 of the SCIP Optimization Suite is now available. This brings updates to GCG, PaPILO, SCIP, and SoPlex: - github.com/scipopt/gcg/... - github.com/scipopt/papi... - github.com/scipopt/scip... - github.com/scipopt/sopl... #orms

github.com

gcg/CHANGELOG at v371 · scipopt/gcg

Contribute to scipopt/gcg development by creating an account on GitHub.

0104

Rolf van der Hulst @rolfvanderhulst.bsky.social · 13/12/2024

Are these talks recorded and posted anywhere so that I can watch them back later?

110

Reposted by Rolf van der Hulst

Sophie Huiberts @sophie.huiberts.me · 10/12/2024

I recorded my current talk for youtube! If you are interested in linear programming or theoretical computer science, this is for you

youtu.be

Open Problems & Diet Problems

YouTube video by Sophie Huiberts

34210