Emtiyaz Khan (@emtiyazkhan) 's Twitter Profile
Emtiyaz Khan

@emtiyazkhan

Team leader at @RIKEN_AIP_EN. Opinions my own. Follow me at mastodon.social/@emtiyaz#

ID: 927837253

linkhttps://emtiyaz.github.io calendar_today05-11-2012 15:32:22

5,5K Tweet

10,10K Followers

228 Following

Vincent Fortuin @vincefort.bsky.social (@vincefort) 's Twitter Profile Photo

We're finally going with the times and are using OpenReview for #AISTATS2025, so you have no excuses anymore to not submit all your great papers!

Kai Arulkumaran (@kaixhin) 's Twitter Profile Photo

Thanks for the invite! The story behind this slide is that I was aware of Emtiyaz Khan's sharp eye for technical questions, and tried to avoid them by keeping most things high-level And then he pointed out "Disclaimer #2" had a number in it ⚰️

Emtiyaz Khan (@emtiyazkhan) 's Twitter Profile Photo

We have two open post-doc positions. You dont' have to be a Bayesian but somebody who is interested to work with at the intersection of DL, Bayes, and optimization. riken.jp/en/careers/res… Interest in understanding deep learning and continual lifelong learning is a plus!

Sarath Chandar (@apsarathchandar) 's Twitter Profile Photo

This is an excellent opportunity for students looking for postdoc positions in continual learning / deep learning / optimization. Emtiyaz Khan is a thoughtful leader with whom you would love to work!

Simone Scardapane (@s_scardapane) 's Twitter Profile Photo

*Model Merging by Uncertainty-Based Gradient Matching* by Nico Daheim Thomas Möllenhoff Edoardo Ponti Iryna Gurevych Emtiyaz Khan Adding Hessian preconditioning improves model merging, and it can be justified in terms of gradient mismatch wrt a multitask model. arxiv.org/abs/2012.03351

*Model Merging by Uncertainty-Based Gradient Matching*
by <a href="/ndaheim_/">Nico Daheim</a> <a href="/tmoellenhoff/">Thomas Möllenhoff</a> <a href="/PontiEdoardo/">Edoardo Ponti</a> <a href="/IGurevych/">Iryna Gurevych</a> <a href="/EmtiyazKhan/">Emtiyaz Khan</a> 

Adding Hessian preconditioning improves model merging, and it can be justified in terms of gradient mismatch wrt a multitask model.

arxiv.org/abs/2012.03351
Stephan Mandt (@stephanmandt) 's Twitter Profile Photo

Stressed out about not meeting the ICLR 2026 deadline?😪No worries! Submit your papers to AISTATS Conference and enjoy Thailand! 🌴😃 Abstract registration: October 3 Full paper submission: October 10 Emtiyaz Khan, Shipra Agrawal, yingzhen

Emtiyaz Khan (@emtiyazkhan) 's Twitter Profile Photo

Videos of talks held during the 2nd Bayes-duality workshop in June 2024 are now publicly available. Playlist: bayesduality.github.io/workshop_2024.… Each talk separately: bayesduality.github.io/talks_2024.html

Thomas Möllenhoff (@tmoellenhoff) 's Twitter Profile Photo

Are you LoRA fine-tuning LLMs and looking for easy ways to get improvements in accuracy? And also Bayesian uncertainty on top for free? Then check our recent work, accepted NeurIPS24 Workshop: Fine-tuning in ML workshop! arxiv.org/abs/2411.04421

Emtiyaz Khan (@emtiyazkhan) 's Twitter Profile Photo

I will be deleting this account soon! Please find me in the federated universe. Address in my webpage. emtiyaz.github.io I will miss this place, but it has changed a lot and very quickly. Time to say goodbye to my data in here.

Emtiyaz Khan (@emtiyazkhan) 's Twitter Profile Photo

List of accepted papers for AISTATS 2025 is now available. aistats.org/aistats2025/ Congratulations to the authors and thanks to the reviewers, AC, and SACs for their help. Thanks to my co-chair Shipra Agrawal & workflow chairs: Christopher Anders (RIKEN) & Tingting Ou (Columbia).

Ruairidh Battleday (@rmbattleday) 's Twitter Profile Photo

🔊 🔊 One month left until our Global Summit on Open Problems for AI (Tokyo, 9-11th April): algopreneurship.ai Come and join Tokyo's AI leaders to discuss the next 5 years in AI Algorithms, Ethics, and Applications. Program released and discounted student, scientist, and

🔊 🔊 One month left until our Global Summit on Open Problems for AI (Tokyo, 9-11th April): algopreneurship.ai

Come and join Tokyo's AI leaders to discuss the next 5 years in AI Algorithms, Ethics, and Applications.

Program released and discounted student, scientist, and