I will be an Area Chair of NeurIPS 2024. My plan to serve our ML community the way it deserves:

Do my best to find the most relevant reviewers for each paper.

Make sure reviewers engage properly with authors.

Read each rebuttal and comment to make informed decisions.

Curious to unpack how acceleration in training impacts generalization: 👇👇👇

REMINDER: 13 PhD positions in Norway

DEADLINE: Monday March 4

APP. LINK: jobbnorge.no/en/available-j…

Start your interdisciplinary PhD in with engaged supervisors and many fellow students.

Projects: integreat.no/about/vacancie…

Journal of Machine Learning Research Transactions on Machine Learning Research NeurIPS Conference ICLR 2024 ICML Conference

The Proceedings of NLDL 2024 are now available online on proceedings.mlr.press/v233/

🚨 NLDL Conference 2024 proceedings are public

🎉🎉🎉

If you have questions regarding 13 PhD Positions at Integreat - Norwegian Centre for Knowledge-driven Machine Learning, do not miss this open zoom meeting happening in a few minutes:

integreat.no/events/public-…

RobustTailor formulates a realistic game between an attacker and a server each equipped with a set of poly-time runnable attack/defense programs and optimize their actions by simulation with regret bounds. The game framework can be applied in practice beyond ML Security ...

Servers in federated learning are vulnerable. Despite numerous robust aggregators proposed, none are entirely immune to attacks. By framing the robust learning problem as a strategic game between an adversary and a server, RobustTailor boosts existing aggregators via simulation.

#TMLR “Mixed Nash for Robust Federated Learning” introduces RobustTailor, a simulation-based framework boosting aggregators.

Designing an omnipotent defender is tough. Instead, utilize existing robust aggregators.

openreview.net/pdf?id=mqMzerr…

w/ Thomas Pethick Ali Ramezani Volkan Cevher

We have 36-month Postdoc Position in ML (Salary NOK 575 400 – 657 300):

Joint Physics-informed and Data-driven Complex Dynamical System Solvers

Reach out to me if you want to apply for this position.

jobbnorge.no/en/available-j…

For PhD candidates, one great paper (even one ongoing but 'super' draft) works better than several low-quality papers.

mn.uio.no/ifi/english/re…

The code for RoSA: Accurate Parameter-Efficient Fine-Tuning via Sparse + Low-Rank Adapters is now available:
github.com/IST-DASLab/RoSA
along with a PEFT integration
github.com/IST-DASLab/pef…
As a bonus, we also QRoSA, which implements the same idea, but with quantized base weights.

Super glad to be a panelist at AI Industry Week 10-12 April Telenor Arena & Online. Will talk about AI Cybersecurity Research & Science.

Do not miss this great event with major industries in AI and cybersecurity!

aiindustryweek.no/#speakers