Researcher in math. optimization @inria_grenoble. Applied maths, programming, OSS, energy, higher ed, games. Also on the elephant and blue sky platforms
Berlin, Germany
Joined December 2012
- Tweets 5,194
- Following 2,130
- Followers 1,723
- Likes 14,247
Mathieu Besançon retweeted
In our new work, with @matbesancon, we study #robustoptimization problems from an oracle perspective. We develop #FrankWolfe type algorithms applied to a smoothed version of the problem, which converge to an optimal solution within a certain number of oracle calls. 1/n
The unique @giomdal making your code run faster than a TGV at the Julia Optimization days in Paris at CNAM
This tweet is unavailable
Patrick Gelß from @ZuseInstitute continuing the @FC_MODAL @RIKEN_AIP_EN workshop with tensor decompositions, and showing us how to represent graphically tensors of higher order as stickmen
Have been trying @obsdmd for about a week now, very impressed with the tooling, structure, onboarding for research. Every single tool I tried usually died out way before
I hesitated last year when I took mine, very happy to have gone with a company centered on selling domains on that one
Very hyped to see friends and colleagues for three interdisciplinary days on optimization, ML, and physics at the @ZuseInstitute
This tweet is unavailable
Finally some hope to have. Now policy makers should build infrastructure up to these expectations
"Trains, not planes, are increasingly business travellers' preferred choice in Europe. Higher airfares and lower carbon emissions behind a significant shift to trains." skift.com/2023/04/17/trains-…
That feeling when you need a super-specific feature from an algorithm, brace yourself to modify a library implementing that algorithm to get out what you want... and realize the feature is already there from the start
Big shoutout to github.com/JuliaGraphs/Graph…
Just added the @TheSIAMNews Summer School on Quantum Computing and Optimization (July 30th, at @LehighU)
The start of the list is up, ping me to add yours!
matbesancon.xyz/post/2023-01…
No delay, a pile of papers, breakfast and coffee. Despite the "surprises", train travel remains very pleasant, let's make it more reliable, affordable, and flexible
Sébastien spearheaded the work, this also opens promising lines, leveraging computational optimization for quantum information problems
Trivia: it's also the first time I co-author with another French researcher
Very fun work, quite some intensive number crunching at the limit of the capabilities of current technologies (had been a while since I saw the usual MIP & LP solvers all crash on a problem!)
Finally, we certify the separating hyperplane with an exact quadratic unconstrained binary optimization (QUBO) formulation, which has gained quite some traction in the last years
The catch: optimizing over the local polytope is NP-hard, but we can use heuristics providing us extreme points in a provably convergent Frank-Wolfe approach to derive a proof of membership or a separating hyperplane
New preprint fresh from the oven with colleagues from the @ZuseInstitute to start the weekend:
we tackle the membership/separation problem for the Bell local polytope in quantum information
arxiv.org/abs/2302.04721
The start of the list is up, ping me to add yours!
matbesancon.xyz/post/2023-01…
@thserra do you know if someone picked up your excellent blog post idea on summer schools for 2023? Now that they are starting in person again?