Jan Stühmer @Neurips

Great first day of the 3rd annual CHAIR Structured Learning Workshop @ Chalmers! 🥳

Event page & agenda: ui.ungpd.com/Events/60bfc...

1st day featuring:
@betapata.bsky.social
@janstuehmer.bsky.social
@arnauddoucet.bsky.social
@frejohk.bsky.social

4/4 We believe that our proposed Clifford Frame Attention can also be suitable for other protein structure related machine learning models, as a drop-in replacement for Alphafold's IPA.

3/4 The proposed model achieves high designability, diversity and novelty, while also following the statistical distribution of secondary structure elements found in naturally occurring proteins.

2/4 We introduce a generative model for protein backbone design that utilizes geometric products and higher order message passing and propose Clifford Frame Attention (CFA), an extension of AlphaFold's invariant point attention (IPA).

Take a look at our #neurips2024 paper, "Generating Highly Designable Proteins with Geometric Algebra Flow Matching" which at the same time more closely follow the distribution of secondary structure elements

arxiv.org/abs/2411.05238

Project page: www.h-its.org/projects/gafl/

#Ai4Science

#ai4science #machineleaerning

I have created a starter pack for AI for Science. Let me know if you would like to be added.

go.bsky.app/FPgU6Pt

Generating Highly Designable Proteins with Geometric Algebra Flow Matching We introduce a generative model for protein backbone design utilizing geometric products and higher order message passing. In particular, we propose Clifford Frame Attention (CFA), an extension of the...

Generative flow matching model for highly designable proteins:

Simon Wagner & Leif Seute et al., Generating Highly Designable Proteins with Geometric Algebra Flow Matching, Neurips 2024
arxiv.org/abs/2411.05238

Graph Neural Networks and Spatial Information Learning for Post-Processing Ensemble Weather Forecasts Ensemble forecasts from numerical weather prediction models show systematic errors that require correction via post-processing. While there has been substantial progress in flexible neural network-bas...

Calibrated predictions with GNNs:

Moritz Feik et al., Graph Neural Networks and Spatial Information Learning for Post-Processing Ensemble Weather Forecasts, Earth System Modeling Workshop @ ICML2024,
arxiv.org/abs/2407.11050

Grappa -- A Machine Learned Molecular Mechanics Force Field Simulating large molecular systems over long timescales requires force fields that are both accurate and efficient. In recent years, E(3) equivariant neural networks have lifted the tension between co...

Learnable parameterized force field models:

Leif Seute et al., Grappa - A Machine Learned Molecular Mechanics Force Field, AI for Science Workshop @ ICML2024 arxiv.org/abs/2404.00050

Let me highlight some of our papers we published last year thanks to our amazing Bachelor, Master and PhD students: