Visiting Assist Prof @MITSloan Assist Prof @JohnsHopkinsAMS, Optimization PhD from @Cornell_ORIE Here to share pretty maths/3D prints/research
Baltimore, MD
Joined January 2015
- Tweets509
- Following448
- Followers3.3K
- Likes2.1K
Ben Grimmer retweeted
(1/9)We are pleased to announce the Optimization × AI Forum, a new online seminar series on how AI is changing the way optimization research is done.
Progress in AI for mathematics and science is now so rapid that workshops, conferences and arXiv postings can no longer keep pace.
Current optimistic feeling on the future of my research:
(1) Theory at its best can lead practice,
(2) Theory is strapped to an exponential rocket in AI,
(1+2) Now is the best time to meaningfully work in theory
(At least in my domain of mathematical optimization)
Ben Grimmer retweeted
Exact Theory and Optimal Stepsizes for Stochastic Gradient Descent i...
arXiv:2610.42048
Stochastic Rayleigh-Ritz Ascent: Global Linear Convergence under a S...
arXiv:2610.42071
Just appeared on arXiv, we derived exact theory and minimax optimal stepsizes for Stochastic Gradient Descent (SGD)'s last iterate.
For smooth convex minimization, our theory comes with exactly matching lower bounds proving no further progress is possible.
Below are the unique optimal, nearly linear stepsize schedules for smooth convex stochastic optimization. Our Corollary 1.2 gives the formal result.
For those interested in reading more, please see the link to arxiv below :)
Our paper, only 14 pages long and the analysis is quite short and sweet:
arxiv.org/abs/2610.42048
On my way to INFORMS 2026 in San Francisco, will be around all week.
For those interested and also around, I'll be presenting some work on subgame perfect gradient methods on Monday at 11am in Moscone South-212.
Ben Grimmer retweeted
Congratulations Luhao Zhang and Jincheng Yang on receiving the 2026 @INFORMS Optimization Society Young Researchers Prize!
Their research aims to improve decision-making under uncertainty, with applications in supply chains and AI.
engineering.jhu.edu/ams/news…
Ben Grimmer retweeted
Its been some time since I've read a really nice optimization paper:
arxiv.org/abs/2104.11179v5
In short - a beautiful theory for transforming arbitrary constrained problems into equivalent unconstrained ones with a Lipschitz objective.
This week, I am visiting Brown (Monday) and Cornell (Friday), speaking about "Optimization in the Space of Algorithms",
Discussing metaquestions about how to enumerate all optimal first-order algs or how to optimally adjust an alg at runtime.
If you are in the areas, let me know
Ben Grimmer retweeted
According to my memory from last night and the wayback machine, OpenAI updated the Navier-Stokes pdf at Tue, 08 Sep 2026 19:09:35 GMT. The original version was one page shorter and did not contain any citations to Diego Córdoba and Luis Martínez-Zoroa:
web.archive.org/web/20260000…
Ben Grimmer retweeted
The Silver Rate Is Tight for GD!
After non-trivial interaction with ChatGPT 6, we prove the lower bound n^{-(p+C\sqrt{\log\log n/\log n})}, p=\log_2(1+\sqrt{2}), for GD with predetermined nonnegative stepsizes in smooth convex optimization. The silver exponent is optimal.
Wanted to spotlight some very neat math by Kevin Shu and Alex L. Wang:
Their recent preprint gives teo explanation of so-called "H-dualities". These are (magical imho) relationships between algorithms where reversing the order of stepsizes used provides new meaningful methods
(2) Whenever H dualities hold, a simple algebraic process converts the association "PEP" proof.
The analysis behind these insights is pretty slick, working equally for gradient methods and fixed point operators. It's a fun read for anyone interested in mathematical optimization
Link for those interested: arxiv.org/abs/2609.03281