I talked about decentralized bilevel optimization.

When training a DNN model (specially LLMs) we often clip the gradients. Why is that? Something to do with exploding gradients? We Fabian Schaipp Guillaume Garrigos Umut U. Simsekli have a story to tell on how clipping is computing the geometric median of gradients. arxiv.org/pdf/2402.12828…

The INFORMS Optimization Society conference early check in happening now. Look forward to seeing everyone!

This is pointless and contributes only to worsening hype about AI, and the college admissions rat race.

Our SIAM Review paper is online today. This paper is an extended version of our 2020 SIOPT paper which proposed the ManPG algorithm and was chosen for the SIAM Review SIGEST Award. Congratulations to coauthors Shixiang Chen, Anthony So and Tong Zhang. epubs.siam.org/doi/abs/10.113…

Happy to have this paper on 1-sample zeroth-order Riemannian opt accepted in SIAM Journal on Optimization: arxiv.org/abs/2309.14506 Includes generic approx. bounds between costly 'parallel transports' and cheaper 'vector transports' Joint work with Jiaxiang Li and Shiqian Ma

The website VFS Global @vfsglobalcare for booking France visa appointments is not functioning properly; this time, I can't receive the OTP.... Over the past four months, I've encountered numerous issues with this website. This is the worst service I've ever experienced, and it

Just spoke with someone from @vfsglobalcare on the phone and was told their system is not compatible with any Apple device. They suggested I use Chrome and clear all cookies. Seriously? This service is terrible.

At #ISMP2024 , today at 2pm in 510C, I will talk about adaptive Barzilai-Borwein method. It is a line-search-free, parameter-free gradient method, a very simple modification to the BB method. We prove the O(1/k) convergence rate for general convex functions.

departed from Montreal at 8:30am. 3 hours later, we are back to Montreal due to technical errors with the aircraft. Traveling experiences in this summer have been not good.

I have a postdoc opening on continuous optimization and machine learning at Rice University. Interested candidates please send a CV to Shiqian Ma at [email protected]. Review of the applications will start on 09/23. Please help spread the word. Thanks!