Diffusion modeling from a Bayesian inference perspective: instead of denoising a sample, we infer it from repeated noisy measurements. Our perspective generalizes Bayesian Flow Networks and achieves better sample quality in fewer steps than BFNs. arxiv.org/abs/2502.07580 github.com/martenlienen...

Real data is noisy but HiPPO assumes it's clean. Our UnHiPPO initialization resists noise with implicit Kalman filtering and makes SSMs robust without architecture changes. Learn more at our #ICML poster: Thu 11am E-2409 Paper: openreview.net/forum?id=U8G... Code: github.com/martenlienen...

BioEmu now published in @science.org !! What is BioEmu? Check out this video: youtu.be/LStKhWcL0VE?...

Scalable emulation of protein equilibrium ensembles with generative deep learning

Following the sequence and structure revolutions, predicting functionally relevant protein structure changes at scale remains an outstanding challenge. We introduce BioEmu, a deep learning system that...

science.org

Microsoft Research@msftresearch.bsky.social · last yr.

Today in the journal Science: BioEmu from Microsoft Research AI for Science. This generative deep learning method emulates protein equilibrium ensembles – key for understanding protein function at scale. www.science.org/doi/10.1126/...

I am truly excited to share our latest work with @mscherbela.bsky.social, Philipp Grohs, and @guennemann on "Accurate Ab-initio Neural-network Solutions to Large-Scale Electronic Structure Problems"! arxiv.org/abs/2504.06087

Accurate Ab-initio Neural-network Solutions to Large-Scale Electronic Structure Problems

We present finite-range embeddings (FiRE), a novel wave function ansatz for accurate large-scale ab-initio electronic structure calculations. Compared to contemporary neural-network wave functions, Fi...

arxiv.org

martenlienen.com/blog/what-is... A summary of my basic understanding of SDEs, Ito and Stratonovich integrals that I gathered from the great book Applied Stochastic Differential Equations #math #machinelearning

I just asked #gemini 2.0 flash thinking to do some #math for me and this is the first time that I have seen an #LLM refer to its own thought process / thinking tokens without "realizing" that I would not see those.