🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

New work on dynamic regret with the stellar Andrew Jacobsen, accepted at NeurIPS arxiv.org/abs/2406.01577 We show that the dynamic regret minimization problem with linear losses is completely equivalent to a static regret problem in an extended decision space. The equivalence

Excited to share new work with Idan Attias and Dan Roy, where we determined the minimax regret of sequential probability against arbitrary expert classes in the contextual setting, resolving a long standing open problem that has been studied since Cesa-Bianchi & Lugosi (2006)

Computing Optimal Regularizers for Online Linear Optimization ift.tt/vGXBaby

All you want to know about spectral transformers in one webpage, papers & code: (& we'll try to keep it updated!) sites.google.com/view/gbrainpri…

2024 Wrapped🎁: Won the Rhodes Scholarship (Rhodes Trust)! Presented first-author work at #NeurIPS! Named a Top 10 Harvard Tech Review Innovator! Co-led new Preprint: 2ly.link/22phA Robots! Survived PhD apps! None of this happens without my mentors; grateful beyond words!

Some ideas on a new optimizer from my student Qinzi Zhang: (github.com/ZQZCalin/train…) Early stages, but the empirical results are really promising! Would love to hear any thoughts, either on the empirical side or analysis-wise, and open to collaboration!

Can you solve group-conditional online conformal prediction with a no-regret learning algorithm? Not with vanilla regret, but -yes- with swap regret. And algorithms from the follow-the-regularized leader family (notably online gradient descent) work really well for other reasons.

You know you've got a nice problem when its discovered independently. In this case a generalization of our OGD results was independently given by Anastasios Nikolas Angelopoulos Jordan and Tibshirani in their elegant work on gradient equilibrium: arxiv.org/abs/2501.08330 Their paper is great.

We have a new results on arxiv: arxiv.org/abs/2502.11673 :)

Suppose you and I both have different features about the same instance. Maybe I have CT scans and you have physician notes. We'd like to collaborate to make predictions that are more accurate than possible from either feature set alone, while only having to train on our own data.

Wow! arxiv.org/abs/2504.09096

New version of my book on Online Learning! arxiv.org/abs/1912.13213 This is a BIG update, with 43 pages of additional content. Let me list some of the changes. 🧵1/8

I have an opening for a post-doc position: I am looking for smart people with a strong CV in optimization and/or online learning All my ex post-docs (Kwang-Sung (Kwang) Jun, Mingrui Liu, and El Mehdi SAAD) became assistant professors, I'd like to continue this trend 😉 Please share it!

As promised, we put on Arxiv the proof we did with Gemini. arxiv.org/pdf/2505.20219 This shows that the Polyak stepsize not only will not reach the optimum, but it can cycle, when used without the knowledge of f*. Gemini failed when prompted directly ("Find an example where the

Julien Weibel (SIERRA), Pierre Gaillard (Thoth), Wouter M. Koolen (CWI), Adrien Taylor (SIERRA): Optimized projection-free algorithms for online learning: construction and worst-case analysis arxiv.org/abs/2506.05855 arxiv.org/pdf/2506.05855 arxiv.org/html/2506.05855

First position paper I ever wrote. "Beyond Statistical Learning: Exact Learning Is Essential for General Intelligence" arxiv.org/abs/2506.23908 Background: I'd like LLMs to help me do math, but statistical learning seems inadequate to make this happen. What do you all think?

Dynamic Regret Reduces to Kernelized Static Regret ift.tt/6HjPz2C