Full Professor of Mathematical Data Science at @univienna and group leader at @oeaw.
Joined May 2021
- Tweets213
- Following160
- Followers521
- Likes727
Philipp Grohs retweeted
📢 @MScherbela, @GrohsPhilipp, @LeonGerard5 and colleagues present an approach that reduces the computational cost to model and compute crystalline materials, such as graphene or lithium hydride, by a factor of 50 compared with previous work. nature.com/articles/s43588-0…
Philipp Grohs retweeted
Fantastic work scaling up neural network VMC methods like FermiNet and Psiformer to previously impossible systems - you can start to tackle real challenging chemistry at this scale!
I am truly excited to share our latest work with @MScherbela, @GrohsPhilipp, and @guennemann on "Accurate Ab-initio Neural-network Solutions to
Large-Scale Electronic Structure Problems"!
arxiv.org/abs/2504.06087
Philipp Grohs retweeted
I am truly excited to share our latest work with @MScherbela, @GrohsPhilipp, and @guennemann on "Accurate Ab-initio Neural-network Solutions to
Large-Scale Electronic Structure Problems"!
arxiv.org/abs/2504.06087
Philipp Grohs retweeted
🎉 Top rankings also in the subject ranking for #STEM departments of @univienna:
- Physics: #48 out of 500 universities (2022: 301-400)
- Mathematics: #34 out of 500 universities (2022: #27)
Find the [German] press release here: ⬇
Die #univie bleibt die bestplatzierte Uni Österreichs im #ShanghaiRanking! Insgesamt konnte sich die Universität Wien in 7 akademischen Fachbereichen unter den Top 100 weltweit platzieren.👏
medienportal.univie.ac.at/me…
Philipp Grohs retweeted
Learn about How well do we know #AI from @GrohsPhilipp et al. in our research magazine Rudolphina rudolphina.univie.ac.at/en/u…
#AI develops in high speed and it is more than necessary to understand it - on a technical level as well as on a philosophical level. Philipp Grohs from @ds_vienna, @univienna proposes to move towards smaller, interpretable neural networks and to make these more intelligent.
Philipp Grohs retweeted
This day comes a bit earlier than I expected
New York Times article on AlphaGeometry
Geometry was my favorite subject in high school because solving them requires many step of reasoning and planning. Geometry problems however have been difficult for AI to solve. Our Nature paper shows my team’s progress in Geometry (approaching IMO Gold level). Key ideas include theorem-proof synthesis and combining search with neural nets (neuro-symbolic) to do complex multi-step theorem proving.
nytimes.com/2024/01/17/scien…
Philipp Grohs retweeted
Apply for my current (soon to be former) job!
It's a pretty unique opportunity, so please DM me with any questions.
jobs.univie.ac.at/job/Scient…
Philipp Grohs retweeted
Any academic institution could leap ahead in the ML race by building a large cluster and it’s slightly puzzling why they don’t try
Philipp Grohs retweeted
📢 Exciting news #teamtomo! We've updated our FakET preprint presenting a new method to simulate cryo-electron tilt-series, with L. Herrmann, @GrohsPhilipp, and @HaselbachLab. To celebrate, we've created a fun movie-style poster. Can you guess the movie? More in the 🧵.
ALT A movie-style poster presenting the research article instead of the original movie poster it was inspired by.
Philipp Grohs retweeted
The Austrian capital Vienna has been named the world's most liveable city cnn.it/42YDsmN
Philipp Grohs retweeted
#CASFellow Philipp Grohs @GrohsPhilipp @univienna gave an exciting talk about the oportunities and limitations of Deep Learning in the sciences. Many thanks to Philipp for this fantastic insight and @GittaKutyniok for moderating.
#CASResearchFocus "Next Generation AI"
Philipp Grohs retweeted
Just arrived in Kigali for #ICLR2023 🇷🇼
I will present joint work with @GrohsPhilipp and Felix Voigtlaender in which we prove that learning ReLU networks to high uniform accuracy requires an intractable number of samples.
Paper: arxiv.org/abs/2205.13531
🧵(1/6)
Philipp Grohs retweeted
Happy Birthday RICAM! Unser Mathematik-Institut in Linz feiert seinen 20. Geburtstag! Zu diesem Anlass gibt es auch eine Ehrung für Gründer Heinz Engl! 🎉🥳 Er erhält die Goldene Ehrennadel des Landes OÖ!
Philipp Grohs retweeted
Diffusion models 🤝 Optimal control
Join my presentation tomorrow in the LoGG reading group organized by @HannesStaerk and @dom_beaini.
🧵 Paper: xitter.cf/julberner/status/15903… & more to come!
This Monday, @julberner presents his paper "An optimal control perspective on diffusion-based generative modeling" in our reading group! arxiv.org/abs/2211.01364
Join us on zoom at 11am EST / 5pm CET tomorrow: m2d2.io/talks/logg/about/
Co-authors:
@lorenz_richter @karen_ullrich
Philipp Grohs retweeted
A hugely important matter for @Europarl_EN. Clearly, this reinforces the need for publicly funded research that can provide solutions to counter the misuse of technology for undemocratic, disruptive and - frankly - contemptuous means. @EP_Industry @MEP_Ehler ://bit.ly/41uEJT9
Philipp Grohs retweeted
📢How can the 🇪🇺 achieve its ambitions for #strategicautonomy in #digital technologies?🌐
🎯The Guild shares in its latest position paper its priorities for digital and #ArtificialIntelligence research in #HorizonEurope 💡
➡️ bit.ly/3E4dGE4
#EUHaveYourSay
Philipp Grohs retweeted
The algorithm, developed by researchers from Austria, has given the best results yet for molecules including water and benzene
chemistryworld.com/news/deep…
Philipp Grohs retweeted
🎙️Come to our oral today at the #NeurIPS2022 workshop on score-based methods and learn more about the connection between diffusion models and optimal control (2:40-2:50pm, room 293-294) ⏬
Explore the connection between diffusion models and optimal control 🔥
🎙️Come to our oral at the #NeurIPS2022 workshop on score-based methods and let’s discuss how one field can benefit from the other.
📖bit.ly/3UAhena
Great work with @lorenz_richter and @karen_ullrich.
Philipp Grohs retweeted
Great news! Our book on "Mathematical Aspects of Deep Learning" by @CambridgeUP joint with @GrohsPhilipp has now been printed. Please check at cambridge.org/de/academic/su….
@researchbavaria, @ResearchGermany, @AIatLMU, @LMU_Muenchen
Philipp Grohs retweeted
📢📢New feature in #NeuralCompression repo: Bits-Back compression for diffusion models!
Compress image data 🖼️ using diffusion models at an effective rate close to the (negative) ELBO.
See: github.com/facebookresearch/…
Some context ⏩[1/4]
Dear neural compression enthusiasts,
there is a new @PyTorch -repo in town
github.com/facebookresearch/…
Includes;
🔥neural image and video compression
🔥bits-back coders
🔥GPU entropy coders
We already work on extensions, feel invited to contribute ❤️❤️❤️
