Pascal Mettes

Our survey "Hyperbolic Deep Learning in Computer Vision: A Survey" has been accepted to #IJCV! The survey provides an organization of supervised and unsupervised hyperbolic literature. Online now: link.springer.com/article/10.100… w/ @GhadimiAtigMina @mkellerressel @jeffhygu @syeung10

@adn_twitts Perhaps not a perfect match, but scattering networks: biorxiv.org/content/10.648…

I'm looking for a paper (pre deep learning) about learning a filter bank from images. It showed that a data-driven filter bank was possible. It resembled the filters learned by CNNs. Does anyone know what I am talking about? Some older people with pre deep learning memory?

🚀 Yesterday we held the fourth session of our Seminar Series! 🧠Prof. @PascalMettes delivered a talk on "Hyperbolic Deep Learning" He explained that hyperbolic geometry is especially suited for hierarchical deep learning, offering better accuracy with fewer dimensions.

✨PhD vacancy alert✨ Joost Batenburg and I are looking for someone that wants to work on fine-grained visual understanding in #ComputerVision @LIACS Apply here before 20 Feb: careers.universiteitleiden.nl/job/PhD-Candid…

✨Only one week to go! The ELLIS Unit Barcelona is hosting its fourth Scientific Seminar. Join us for the Scientific Seminar on January 28th with a talk by Prof. @PascalMettes on "Hyperbolic Deep Learning". Don't miss out! Register here➡️ eventbrite.es/e/ellis-unit-s…

@CSProfKGD Same here, no need for emergency reviews in my batch. Although verbose, it seems that the messages from the PCs are taken seriously by the reviewers.

#CVPR2026 AC update: 1 missing review, I have a commitment that the review will be submitted. Overall, easiest AC paper batch I've ever handled with most submitted before the deadline & more importantly the reviews are detailed & justified! GREAT job reviewers!

✨ Kind reminder! The ELLIS Unit Barcelona is hosting its fourth Scientific Seminar. Join us for the Scientific Seminar on January 28th with a talk by Prof. @PascalMettes on "Hyperbolic Deep Learning". Don't miss out ➡️ellisbarcelona.eu/ellis-unit-bar…

@2ptmvd @CVPR Yes, one month between submission and reviewer assignment is a huge gap. I wonder whether the openreview issues of this year played a role.

@PascalMettes @CVPR Fair point on the volume, but the timeline was much different! If I recall correctly, 10-15 years ago assignments dropped in early December. That made a huge difference. The current window forces the workload specifically into the holiday break, regardless of the paper count.

The CVPR review window is officially Dec 18 – Jan 8. 🎄 It’s getting worse every year. We’ve gone from a 25-day window last year to 21 days this year—right over the peak of the holidays 🤦‍♂️ @CVPR

I am very excited to visit the ELLIS Unit in Barcelona next month!

@BlackHC @iclr_conf Unfortunately, the latter is a realistic scenario and already happened. Beyond bribing, people can hold grudges, threaten, sabotage papers in future conferences, etc. We care too much about careers instead of science and people end up doing unhealthy things to advance them.

@PascalMettes @iclr_conf pretend nothing happened re reviewers and shuffle the ACs a bit. High quality reviews and discussions aren't affected by this bc they stand on their own. Low quality ones are noisy anyway And I can't believe people would try to blackmail and bribe ACs at a large scale 🥺

Lol what a shitshow @iclr_conf I'm sure the new ACs will take the rebuttals into account in a meaningful way when they decide to keep the original scores. What a waste of effort for everyone who spent time on rebuttals, and what a stupid reaction to the leak 🤦

I have funding for 2 new PhD students on hyperbolic deep learning. Interested in joining this exciting research direction? Check out the vacancy and consider applying! Link: werkenbij.uva.nl/en/vacancies/t…

I will even go as far as to say that the review quality has been stable over the last decade (seen as author and AC). My advice, especially for new students, is to not go to twitter after reviews have been released, too many frustrated people with ill-intended messages. [4/4]

In my experience, most reviews contribute to the process and with 4 reviews per paper, a noise review can be compensated by others. There are still plenty of cases with errors, but this problem is as old as peer review itself. [3/4]

It is shocking to me how much of an echo chamber AI twitter has become regarding peer review. There are 80k reviews for ICLR and people here loose their minds about noise that has always been there. LLM reviews are bad, noise is bad. But this discussion is so un-scientific. [1/4]

@xwang_lk I don't expect anything to be very different, just people on twitter screaming. There are 80k reviews, surely some will be bad. It has always been like that.

What is wrong with ICLR reviewing this year? It used to be reasonable most of time.

@xwang_lk I disagree. Good surveys create order in the chaos of a subfield and are a bridge from the frontier of science to the classroom, which is a non-trivial but undervalued step. Also not nice to judge others for their oeuvre, we all have a role to play in this world.

Regardless of the preprint nature, A LOT of people publish survey or position papers just for citations. IMO, it is actually a shame if your top cited papers are mostly survey or position papers.

@GroundlessHope @jcponcemath @geogebra Very cool site, thanks for the link!

@jcponcemath @geogebra @PascalMettes

Did you know you can do hyperbolic geometry in @geogebra? Check this out by Rafael Losada Liste: geogebra.org/m/vuuufnr8

@Eng_Hemdi Thanks!

@PascalMettes Awesome , congrats !

I am very happy to share that I received the NWO VIDI grant for my research on hyperbolic computer vision. Many more hyperbolic papers on the horizon the next few years! ivi.uva.nl/content/news/2…