Our new work has appeared on arXiv: arxiv.org/abs/2408.05215, which addresses a challenge in #Machinelearning interatomic potentials by introducing a #physics-informed, #weakly supervised approach.
Our approach improves the accuracy of MLIPs applied to training tasks with sparse training data sets and reduces the need for pre-training computationally demanding models with large data sets.
1
1
38
mtmktmkt retweeted
Watch NEC Research Scientist @federico_errica explain why nearest-neighbor #graphs are not a good solution in fully supervised #machinelearning scenarios. youtu.be/KWwajQs--jE. #NECLabs #NeurIPS23
mtmktmkt retweeted
Online LoG conference 27th – 30th Nov!
All the talks will be live-streamed on Zoom and YouTube with poster sessions and sponsor sessions on GatherTown:
Schedule Zoom: logconference.org/
Youtube: youtube.com/@learningongraph…
GatherTown: app.gather.town/app/etG6JBKc…
mtmktmkt retweeted
neural networks cannot magically solve difficult combinatorial optimization problems. a short🧵on the work of Gamarnik pnas.org/doi/abs/10.1073/pna… describing the performance barriers of GNNs. 1/n
mtmktmkt retweeted
How to train uniformly accurate #MLPotentials from scratch? With @DHolzmueller, @henchristiansen, @federico_errica, @f_alesiani, @mtmktmkt, @Mniepert, and @GroupKaestner, we investigate the capability of uncertainty-biased #MD: researchsquare.com/article/r…. 1/
mtmktmkt retweeted
If want to learn and contribute in this area, we have an opening: lnkd.in/ecHeH-wu neclab.eu/join-us
The mix of skills in the team is quite rare: machine learning, chemistry, physics and computational methods. A lot of potential for fun (...and impactful science)
Our advances in #machinelearning can benefit multiple #research areas by improving the accuracy and efficiency of simulations e.g., to accelerate screening for new drug candidates. Learn how: youtu.be/QCslBd4dKXQ. #NECLabs #computationalscience
mtmktmkt retweeted
It's an exciting time where #NLProc research is moving closer & closer to applications🗨🙋🗨🤖My experienced NLP team is looking for support on the applied side to make a real world impact together🚀🌎
Apply here: neclab.eu/join-us #hiring #LLMs #SoftwareEngineering
mtmktmkt retweeted
📢#AI4Science Talk on 28.08 at 15:00 (CEST) / 09:00 (EDT) on “PDE-Refiner: Achieving Accurate Long Rollouts with Neural PDE Solvers” by @phillip_lippe from MSR/UvA.
Please join us on Zoom if you're interested!
Details: ai4sciencetalks.github.io/pr…
#ML4science #CFD #PDEs #AI4Science
Our #ICML2023 poster presentation is today’s morning session at Exhibit hall 1 #522. We look forward to a thrilling discussion!
We are thrilled to announce that our new paper ”Learning Neural PDE Solvers with Parameter-Guided Channel Attention” has been accepted by #ICML2023 !! Draft will be appeared soon.
1
79
mtmktmkt retweeted
Excited to share our latest survey paper: "Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems"! 🚀 ArXiv: arxiv.org/abs/2307.08423.
We are thrilled to announce that our new paper ”Learning Neural PDE Solvers with Parameter-Guided Channel Attention” has been accepted by #ICML2023 !! Draft will be appeared soon.
mtmktmkt retweeted
We are happy to introduce you to our first MLSS^S speaker: @Mniepert, Professor at @Uni_Stuttgart!
👉 Make sure to register to MLSS^S: the applications are open until 8th April: mlss2023.mlinpl.org/registra…
mtmktmkt retweeted
We are kick starting 2023 with a new invited talk series on #AI4Science #ML4Science
As a first talk, we're excited to have @simonbatzner from @Harvard who will give a presentation about his recent work on molecular dynamics simulations, specific. NequIP (github.com/mir-group/nequip)
mtmktmkt retweeted
Now that #ICML is asking area chairs to make a statement of “why they think peer review helps science?” It is time to debate whether that is actually true. I personally align with what is said in the article below that I recommend. TLDR: Peer review is an experiment that failed!
Replying to @a_m_mastroianni
Original post here (please don't fire me 🙂): experimentalhistory.substack…
Thank you very much for your interest in our PDEBench! We are going to have a hard work to improve it👍
PDEBench will be presented in today’ morning poster session at #NeurIPS2022 . Please feel free to visit our poster at Hall J #1030, and ask any questions and have a discussion with us!!




















