Nicolas Deutschmann
@nickgermann
Followers
236
Following
633
Media
40
Statuses
533
ML Researcher @cradlebio GenAI for proteins with a touch of uncertainty quantification. Views are my own. (he/him)
Zurich, Switzerland
Joined November 2007
Our paper got accepted at #AAAI24! 🎉.We propose two new .sequence generation algorithms with "error bars", by adapting beam search to conformal predictions. If you use LLMs for science, like predicting molecules or proteins that verify some conditions, check it out!.
Here's our preprint on conformal sequence generation!.w/ @malberts99 @MariaRoCompBio. Two new algorithms for sequence prediction sets with guarantees:.- Conformal sub-beams: pruning standard beams.- Conformal dynamic beams: step-by-step conformal decoding.
0
8
26
RT @nathanbenaich: Zurich AI meetup speakers and friends for 4 Mar!. - Peter Kontschieder, Research Director at Meta. We met back in Mapill….
0
4
0
RT @MilesCranmer: Nice write up by Kyle!. I’ll add on - answering questions like "why do we exist" needs no industrial application! It is t….
0
3
0
RT @PatrickKidger: ⚡️ New blog post. whilst wearing my Cradle hat!! 🧬. "State-of-the-art enzyme engineering with fully-automated GenAI"….
0
5
0
RT @PatrickKidger: ✨Aaaand it's time for ICML!. DM me or @constance_frg if you'd like to chat about 🧬ML-for-proteins or 🚀startups! :D https….
0
3
0
RT @jelleprins: 🔥 We are hiring engineers in Amsterdam!. Come build the future with us: a web app to design tiny, biological machines, that….
0
10
0
RT @bourmeau: Si nous étions en Grande-Bretagne ou aux USA, une telle information ne se serait pas enfouie dans un live mais ferait le gros….
0
2K
0
RT @marianna_raps: 📢 We are hiring! 📢. Are you a MSc/PhD👩‍🎓👨‍🎓 passionate about #AI #ML in #cancer, looking for your next career step?. Joi….
0
82
0
RT @marianna_raps: Join our Track on #AI for #drugdiscovery at #AMLD2024 March 25 in Lausanne 🇨đź‡!.🔥 Sessions on #AI for #molecules, #single….
0
3
0
RT @malberts99: Interested in Uncertainty Quantification for Sequence Prediction?.Come check out our poster at @RealAAAI today. We add rigo….
0
12
0
RT @AleksandrosSob1: We have a new preprint on DFT and eigenvalue algorithms:."Hermitian Pseudospectral Shattering, Cholesky, Hermitian Eig….
arxiv.org
Approximating invariant subspaces of generalized eigenvalue problems (GEPs) is a fundamental computational problem at the core of machine learning and scientific computing. It is, for example, the...
0
5
0
This is so cool! I keep pushing back using JAX thinking “so much of my current work is in PyTorch”. Can’t wait to try it out!.
Announcing torch2jax! Run PyTorch code natively in JAX. 🤝. Mix-and-match PyTorch and JAX code with seamless, end-to-end autodiff, use JAX classics like jit, grad, and vmap on PyTorch code, and run PyTorch models on TPUs.
0
0
5
RT @malberts99: Excited our Preprint on Conformal Beam Search with @nickgermann and @MariaRoCompBio is out on @Arxiv. We add rigorous uncer….
0
4
0
One of the cool surprises from this work is how well calibrated the set sizes turned out to be in our experiments with dynamic beam sizes:.Great correlation between how wide the beams need to be vs how they actually are. Excited to see how well this holds in other exps!.
This is relevant for any problem where an autoregressive language model is tasked with generating an exact or approximately exact sequence. Our algorithm generates prediction sets with a guarantee (e.g. 99%) that the correct sequence is present in the set!
1
2
9
Here's our preprint on conformal sequence generation!.w/ @malberts99 @MariaRoCompBio. Two new algorithms for sequence prediction sets with guarantees:.- Conformal sub-beams: pruning standard beams.- Conformal dynamic beams: step-by-step conformal decoding.
1
18
57
RT @av_thakkar: Great to see our work @IBMResearch on prompt driven retrosynthesis in @ACSCentSci highlighted by @OPRD_ACS alongside severa….
pubs.acs.org
Data-driven approaches to retrosynthesis are limited in user interaction, diversity of their predictions, and recommendation of unintuitive disconnection strategies. Herein, we extend the notions of...
0
13
0
RT @johannbrehmer: Are you dealing with geometric data, be it from molecules or robots? Would you like inductive biases *and* scalability?….
0
82
0
RT @teodorolaino: Can’t wait to meet new and old connections at #ACSFall2023. If you are around, pass and say 👋🏻 by the @IBMResearch booth….
0
4
0
RT @teodorolaino: On Wednesday, @malberts99 will present at 08:45am:.Automatic structure elucidation from IR spectra . and at 04:15pm:.Aut….
0
3
0