Mathematical optimization enthusiast and part of the R&D team at MOSEK ApS.

Joined April 2016
11 Photos and videos
Follow me on BlueSky where I have started tweeting. A link to my profile can be found in the replies.
1
34
Nice to see developments in this field - great math. Not sure the benchmark is all that reliable though, as they report 60 ms projection times, as opposed to the 6 μs I was observing. That is a 10'000x difference on similar hardware!?
Luis M. Brice\~no-Arias, Crist\'obal Vivar-Vargas: Projection onto cones generated by epigraphs of perspective functions arxiv.org/abs/2411.08000 arxiv.org/pdf/2411.08000
89
Henrik A. Friberg retweeted
MIPLIB 2024 submissions open: miplibsubmissions.zib.de/ Submit all instances you can share! Closes: 30.11.2024 Restrictions: Mixed-integer linear programs (includes integer programs indicator / SOS constraints).

9
18
1,750
Getting ready for euro2024cph.dk, writing a paper to elaborate a note in my slides: optimization-online.org/2024… Hope to see you all!

1
9
379
Sound propagates at around 300 m/s. Light propagates at around 300 m/μs. (air speeds near earth; 1 significant digit of precision; beautiful and memorable)
1
102
Let child 1 decide on a fair cut into n pieces and let children 2,…,n pick their favorites. Child 1 gets the leftover, and the other n-1 pieces are put back in the middle. Now it’s child 2 to cut and get the leftover, and so it repeats until the utility has been distributed.
1
2
225
Henrik A. Friberg retweeted
17 Jan 2024
Nice post by Ken Fordyce on Linear Programming, shared on the INFORMS Open Forum and partially inspired by Bob Bixby's s.t. episode: connect.informs.org/discussi…
3
15
1,227
Henrik A. Friberg retweeted
No offence to C developers, but why? 🥲
Community note
The source of this claim is Uncyclopedia, a satire website. It seems C has no official mascot. en.uncyclopedia.co/wiki/C
105
151
1,692
225,838
Henrik A. Friberg retweeted
Kicking off 2024 with our paper publishing online (pubsonline.informs.org/doi/1…) in Operations Research! Check it out! @oktaygunluk @69alodi
2
2
50
2,732
I am always happy to hear when my research is being put to use. Thank you for the kind words ☺️
You can found there a Python implementation of the projection onto the exponential cone (H. Friberg, docs.mosek.com/whitepapers/e…). The one proposed in the article is implemented in Julia. Hoping that it can help someone who found himself with the same need. github.com/ozekri/projection
2
3
1,517
Squareplus would also be faster in conic optimization problems as its is one quadratic cone: (2t-x, x, 2) \in Q^3 compared to two exponential cones (see docs.mosek.com/modeling-cook…).

27 Apr 2021
Why do people use softplus? Calling two transcendental functions just to get a non-negative number with smooth gradients seems like overkill. Here's a simple all-algebraic alternative ("squareplus"?) that's twice as fast and seems to do the job.
1
4
613
I wonder if there exists a problem transformation such that this new concept of Polar Convex Programming (naturalspublishing.com/files…) can be solved using traditional convex programming techniques?

1
3
164
My scam detector is on red alert, but who doesn't want such emails to be true? "You have been provisionally selected for the Best Researcher Award at the upcoming International Research Awards on Mathematics and Optimization Methods". Website: maths-conferences.sciencefat…

1
148
Simplex visualization tool: gilp.henryrobbins.com/en/lat…

1
3
37
4,162
Henrik A. Friberg retweeted
On the 1st/15th of every month, I try to release something cool Today, dogbunnypuzzle.com, a new form of puzzle I invented! Though the design is so simple, it can't possibly be truly new (?) Works great on both mobile and desktop
22
32
202
Henrik A. Friberg retweeted
Replying to @gabrielpeyre
Here's a different view angle with the min in blue. It would be interesting to see how global solvers behave on it. I was able to find an SDP-based heuristic here: rdoelman.bitbucket.io/2019/0…
1
1
5