Computational Science and Modelling of materials and molecules at the atomic-scale, with machine learning. Join us in the fediverse @lab_COSMO@fediscience.org
Lausanne, Switzerland
- Tweets 1,148
- Following 450
- Followers 3,720
- Likes 1,130
COSMO Lab π¨ππͺπΊ retweeted
Last week, while we were all busy at #psik2025, my paper with Hanna TΓΌrk got published in @PRXenergy. We used @lab_COSMO's #PET models to study the dynamic reconstruction of LPS surfaces, and how it impacts their structure, stability, and reactivity.
journals.aps.org/prxenergy/aβ¦
1
3
12
673
COSMO Lab π¨ππͺπΊ retweeted
β¨preprint alert: "learning long-range representations w/ equivariant messages" in which we get into the fray of long-range MLIPs and propose to use equivariant charges w/ classical electrostatics as long-range MP mechanism. we design LOREM, an equivariant MLIP, around this...
2
8
36
3,646
Just to bring some π§βπ science fun to those of us still hanging out in this pit of despair, just wanted to share the exciting news you can now run MD (better, FlashMDβ‘οΈ) with up to 64ps timestep for systems as diverse as ala2, Al(110), LiPS, and whatever comes to your mind.
1
3
25
1,596
Check out the #preprint on arxiv.org/abs/2505.19350 and the π§βπ³π recipe atomistic-cookbook.org/exampβ¦ #compchem #machinelearning #faster #better
1
511
π’For those still hanging out here, a new #cookbook recipe landed, demonstrating >2x speedup of *conservative* MD for the PET-MAD universal #MLIP using multiple time stepping. Check out the details at the other place, or head to atomistic-cookbook.org/exampβ¦ #compchem #machinelearning
2
22
916
π¨help needed!π¨ Does anyone, ORB-affiliated or not, can share with us the model weights orb-v2-20241011.ckpt? We need urgently to reproduce some results we obtained with that model checkpoint, but it was removed from their huggingface repo because "the model reached EoL" π€·. π
5
3
1,472
π€« you can get a better universal #machinelearning potential by training on fewer than 100k structures. too good to be true? head to arxiv:2503.14118, to the atomistic cookbook, or to the better science social if you want to find out more about PET-MAD. π§βπ over and out π
1
14
47
3,635
COSMO Lab π¨ππͺπΊ retweeted
Happy new year everyone! After more than three years of work, and splitting out a whole separate package, I am extremely happy to announce the first full public release of featomic, a package to compute representations for atomistic machine learning! 1/n
1
6
24
2,277
Join the dark side of the forces (or not!). TL;DR - Using forces that are not conservative is not a great idea. Latest and greatest, now on the @arxiv , from Filippo Bigi and @marceldotsci.
1
10
30
3,727
There's some nuance though. For that, head to the preprint arxiv.org/html/2412.11569v1 or head to the detailed 𧡠at the better place bsky.app/profile/did:plc:ub6β¦
2
610
torch-pme has landed. if you're curious, head to the better place to see why we're so excited about it π
3
24
1,227
Hello! I'm posting this both here and on the better place, let's see where it gets more re-posts π. We are looking for a research software engineer to help us develop (even) better code for #atomicscale #machinelearning. Check out the specs and apply!
epfl.ch/labs/cosmo/index-htmβ¦
2
9
17
2,935
This is funded by the nccr-marvel.ch, and will involve working together with @corminboeuf_lab and Anirudh Natarajan's lab. Plus, you'll be working with the one and only π§βπ @__luthaf__ so you really should not sit this one out!
2
410
It is time for jack-o-molecule π! Happy chemiscoping chemiscope.org/?load=https%3β¦
1
2
19
1,358
Kind of bittersweet that it's so easy to lean equivariance to the point where it does not matter, but that's the evidence π€·. Also, if you see similar stuff, there's a good set of benchmarks to stress test your unconstrained model!
Great new work by @marceldotsci @spozdn @MicheleCeriotti @lab_COSMO @nccr_marvel @EPFL_en - 'Probing the effects of broken symmetries in #machinelearning' - iopscience.iop.org/article/1β¦ #compchem #materials #statphys #molecules #compphys #simulation #atomistic
1
1
22
2,539
Simple, cheap, and surprisingly accurate. What's not to like of uncertainty estimation from repurposed prediction rigidities?
Great new work by Filippo Bigi @SanggyuChong @MicheleCeriotti @FedericoGrass @lab_COSMO @EPFL_en @UNIMORE_univ - 'A prediction rigidity formalism for low-cost uncertainties in trained #neuralnetworks' - iopscience.iop.org/article/1β¦ #machinelearning #statphys #Bayesian #compchem #AI
2
16
1,521
Hot topics require hot thermostats π‘οΈ
New version of i-PI is out: faster and ready for the machine-learning era in atomistic simulations. New features enabling research on the hottest topics in physical chemistry make the code complete! aippub.org/47O32Pl
8
909
COSMO Lab π¨ππͺπΊ retweeted
accepted now in @MLSTjournal! final manuscript, scripts, and data coming soon. iopscience.iop.org/article/1β¦
well⦠@spozdn and @MicheleCeriotti and I had a look: arxiv.org/abs/2406.17747! In this, we check what exactly non-invariance breaks in practical terms, for water and a PET model. To my surprise, even though angular momentum is not conserved, observables seem largely unaffected!
6
23
2,676
There's lots of choice out there of for algorithms to perform simulations at constant-temperature π‘οΈ. If you want to make an informed choice, there's now a #cookbook recipe ready to try π§βπ³ atomistic-cookbook.org/exampβ¦
1
13
57
4,295
From kinetic temperature to velocity-velocity power spectrum, you can compare how different thermostats fare in achieving equilibrium, sampling slow molecular motions, and preserve dynamical properties.
1
4
361
