Adam Foster

I am very happy to share Orbformer, a foundation model for wavefunctions using deep QMC that offers a route to tackle strongly correlated quantum states! arxiv.org/abs/2506.19960

Last call - open til Oct 30! Are you excited about #MachineLearning and developing new architectures for Molecular Biology? Joint us for the next chapter of BioEmu at @MSFTResearch AI for Science - Berlin DE or Cambridge UK. aka.ms/bioemu-ml

RT @FrankNoeBerlin: Want to join our BioEmu team in @MSFTResearch AI for Science as an Intern? Berlin DE or Cambridge UK are available. Pre…

RT @tom_rainforth: I have an opening for a 2-year postdoc in probabilistic machine learning and/or experimental design. The application de…

I feel fortunate to have been able to contribute to the effort

It's a great achievement for the team and for @FrankNoeBerlin who was at the helm throughout!

@FrankNoeBerlin 👀

RT @fbickfordsmith: There’s a lot of confusion around uncertainty in machine learning. We argue the "aleatoric vs epistemic" view has cont…

Huge thanks to Zeno Schätzle, @PBerntSzab1, @SherryLixueC, @jonkhler, @n_gao96 , Gino Cassella, Jiawei Li, @FrankNoeBerlin and Jan Hermann (2/n)

I am very happy to share Orbformer, a foundation model for wavefunctions using deep QMC that offers a route to tackle strongly correlated quantum states! arxiv.org/abs/2506.19960

Digging into the model we found intriguing behaviour, such as the unsupervised discovery by the model of ‘core’ electron orbitals for second row atoms. This has been a fascinating project to be a part of! Check out the preprint for more details and results. (8/n)

We also saw very strong results confirming the experimental activation energy of a Diels-Alder reaction, and significantly outperforming earlier transferable QMC approaches (7/n)

We scaled this idea up and pushed it to work on strongly correlated systems. On a cost/error plot, we find that Orbformer is on or ahead of the Pareto frontier formed by traditional multireference methods, a first for deep QMC. (6/n)

To get cost down we make use of amortization: solving a single minimization problem with a more complex network that represents multiple wavefunctions simultaneously (5/n)

Describing strongly correlated quantum systems remains a major challenge in quantum chemistry. Deep QMC offer a potential solution, but at a huge computational cost. (4/n)

For a short ML-focused introduction to this paper, I wrote a short blog post ae-foster.github.io/posts/2025/06/… (3/n)

The aleatoric-epistemic view on uncertainty doesn't serve ML researchers' needs and should be replaced. Come to the talk and poster tomorrow (Sat 14 Dec) at the #NeurIPS2024 workshop on Bayesian decisions (gp-seminar-series.github.io/neurips-2024). openreview.net/forum?id=WIjgb…

@j_foerst This rings true. Perhaps because AI research is split over many departments? I knew what happened on the first floor of the stats department, but beyond that I would usually find out outside Oxford. People who moved between departments knew more though

something odd about Oxford (& I wonder if it's the same at other places): I commonly find out about what is happening in AI _at Oxford_ by talking to people _not_ at Oxford. Oxford essentially uses the rest of the world to compensate for the lack of internal communication tools

BIG opportunities to join @MSFTResearch AI for science: one senior researcher and one RSDE position, both focused on applications in molecular biology with the awesome @FrankNoeBerlin. Cambridge, UK or Berlin, DE.

@n_gao96 @NeurIPSConf That's great news and great to see NeurIPS recognising your work on neural net wave functions :) congrats 🎉

I'm thrilled to announce that our work on neural Pfaffians was selected for an oral presentation at @NeurIPSConf 2024! Excited to see neural wave functions gaining more traction at ML conferences!

Interested in working with a highly collaborative, interdisciplinary team to push the state of the art of generative AI for materials design? Join us as an intern by applying through this link! We are the team behind the MatterGen and MatterSim models from Microsoft Research AI for Science. jobs.careers.microsoft.com/global/en/job/…