Announcing the creation of *KAUST Center of Excellence in Generative AI*; official launch on July 1, 2024. Joint with Bernard Ghanem Jürgen Schmidhuber Francesco Orabona David R. Pugh and a few more KAUST colleagues. KAUST funding ($11m over 5 years) + industrial funding. We are looking

When you discuss your own papers in the related work section

The COLT 2024 proceedings are out. Check out some great papers! proceedings.mlr.press/v247/

🚨 New blog post: Dynamic Regret and ADER Dynamic regret and suboptimal path-length bound of Online Mirror Descent. Then, I show how to achieve the optimal dynamic regret bound using ADER by Zhang, Lu, and Zhou,NeurIPS'18 As usual, feedback is welcome! parameterfree.com/2024/07/01/dyn…

Congrats to the best paper award winners at COLT 2024! learningtheory.org/colt2024/award… First up, The Price of Adaptivity in Stochastic Convex Optimization by Yair Carmon and Oliver Hinder:

My student Keyi Chen successfully defended her PhD thesis at Boston University on Generalized Implicit Online Convex Optimization. She did great work on implicit algorithms, like arxiv.org/abs/2306.00201. I am very happy for her!

Career "advice": Live modestly, save money Put your family and health first Keep learning -don't live on your laurels No jerks: don't work for 'em, don't hire 'em

This is scientific malpractice and an intimidation tactic. Senior AI researchers who work in the same area should talk to Guohao Li and get details that they are not sharing publicly (for good reason), and investigate. It’s not a good idea to “let this slide” as it becomes

This is the Way

Twitter just activated a setting by default for everyone that gives them the right to use your data to train grok. They never announced it. You can disable this using the web but it's hidden. You can't disable using the mobile app Direct link: x.com/settings/grok_…

Reading six #NeurIPS24 reviews for the same paper ranging from 8 to 4

Optimization people, how do you call this property? There exists L>0 such that f(y) - f(x*) <= nabla f(y)' (y-x*) - 1/(2 L) ||nabla f(y)||^2 where x* = argmin_x f(x) This is clearly satisfied by a convex L-smooth function, but it is weaker.

As an Action Editor for JMLR and frequent (Senior) AC for ML conferences, I can confidently say there's a noticeable difference in how carefully reviewers are selected. If you have a strong paper and aren't in a rush, consider submitting it to JMLR more often.

I expect the number of ICLR submissions to rise significantly this year, as the China Computer Federation (CCF) has now classified ICLR as an A-level conference.

Every time I teach my Online Learning class, the interactions with the students stimulates new remarks and additions to my notes. This year seems particularly promising 🙂