bkmi

Exploring math and the world.

Benjamin Kurt Miller

My focus is on developing probabilistic models and deep learning algorithms for physics and science. I do this as a Research Scientist at Meta in the FAIR Chemistry team where we work with molecules and materials.

During my PhD I developed work on {simulation-based, likelihood-free, implicit likelihood} inference; a method for solving inverse problems. Applications include gravitational waves, exoplanets, fast ratio bursts, multi-messenger astrophysics, the cosmic microwave background, and dark matter searches.

Timeline

Interned with:

Summary

My PhD was with AMLab and GRAPPA in the AI4Science initiative. Advised by Max Welling, Samaya Nissanke, Christoph Weniger, and Patrick Forré. I am also an ELLIS PhD student working with Max Welling and Gilles Louppe.

I was the primary developer of a simulation-based inference library called swyft, now maintained by Christoph Weniger. I’ve contributed algorithms to simulation-based inference libraries lampe and sbi.

Former member of the Atomic Architects with Tess Smidt. I contributed to e3nn as a developer and used the E(3) equivariant neural network for regression on small organic molecules.

Former member of Artificial Intelligence for the Sciences at the Freie Universität Berlin (Free University of Berlin) with Frank Noé.

Former member of AG Nichtlineare Laserdynamik and Ultrafast AMO Theory. For some time, I worked as a data scientist at a financial tech company.

Here are my relevant social media links: GitHub LinkedIn Twitter and Google Scholar.

Writing

Aaron Havens*, Benjamin Kurt Miller*, Bing Yan*, Carles Domingo-Enrich, Anuroop Sriram, Brandon Wood, Daniel Levine, Bin Hu, Brandon Amos, Brian Karrer, Xiang Fu*, Guan-Horng Liu*, Ricky T. Q. Chen*. Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matching. ICML 2025. 13 July 2025. (Oral at Frontiers in Probabilistic Inference: Sampling Meets Learning. ICLR Workshop 2025).

Chaitanya K Joshi, Xiang Fu, Yi-Lun Liao, Vahe Gharakhanyan, Benjamin Kurt Miller, Anuroop Sriram, Zachary W Ulissi. All-atom Diffusion Transformers: Unified generative modelling of molecules and materials. ICML 2025. 13 July 2025. (Oral at AI for Accelerated Materials Design. ICLR Workshop 2025).

Anuroop Sriram, Benjamin Kurt Miller, Ricky T. Q. Chen, Brandon Wood. FlowLLM: Flow Matching for Material Generation with Large Language Models as Base Distributions. NeurIPS 2024. 10 Dec 2024.

Maximilian Lipp, Benjamin Kurt Miller, Lyubov V. Amitonova, Patrick Forré. Generalizing Coverage Plots for Simulation-based Inference. NeurIPS 2024 submission.

Benjamin Kurt Miller, Ricky T. Q. Chen, Anuroop Sriram, Brandon Wood. FlowMM: Generating Materials with Riemannian Flow Matching. ICML 2024. 21 July 2024.

Benjamin Kurt Miller, Marco Federici, Christoph Weniger, Patrick Forré. Simulation-based Inference with the Generalized Kullback-Leibler Divergence. ICML 2023 workshop on Synergy of Scientific and Machine Learning Modeling. 28 July 2023.

Arnaud Delaunoy*, Benjamin Kurt Miller*, Patrick Forré, Christoph Weniger, Gilles Louppe. Balancing Simulation-based Inference for Conservative Posteriors. 5th Symposium on Advances in Approximate Bayesian Inference. arXiv:2210.06170. 18 July 2023.

Uddipta Bhardwaj, James Alvey, Benjamin Kurt Miller, Samaya Nissanke, Christoph Weniger. Peregrine: Sequential simulation-based inference for gravitational wave signals. arXiv:2304.02035. 04 April 2023.

Benjamin Kurt Miller, Christoph Weniger, and Patrick Forré. Contrastive Neural Ratio Estimation. NeurIPS 2022. arXiv:2210.06170. 27 November 2022.

Benjamin Kurt Miller, Alex Cole, Christoph Weniger, Francesco Nattino, Ou Ku, Meiert W Grootes. swyft: Truncated Marginal Neural Ratio Estimation in Python. Journal of Open Source Software. 19 July 2022.

Wujie Wang, Minkai Xu, Chen Cai, Benjamin Kurt Miller, Tess Smidt, Yusu Wang, Jian Tang, Rafael Gómez-Bombarelli. Generative coarse-graining of molecular conformations. International Conference on Machine Learning 2022. 17 July 2022.

Christoph Hönes, Benjamin Kurt Miller, Ana M. Heras, Bernard H. Foing. Automatically detecting anomalous exoplanet transits. NeurIPS 2021 workshop on Machine Learning and the Physical Sciences. 13 Dec 2021.

Alex Cole, Benjamin Kurt Miller, et. al. Fast and Credible Likelihood-Free Cosmology with Truncated Marginal Neural Ratio Estimation. arXiv: 2111.08030. 16 November 2021.

Benjamin Kurt Miller, Alex Cole, Patrick Forré, Gilles Louppe, Christoph Weniger. Truncated Marginal Neural Ratio Estimation. NeurIPS 2021. arXiv:2107.01214. 02 July 2021.

Benjamin Kurt Miller, Alex Cole, Gilles Louppe, and Christoph Weniger. Simulation-efficient marginal posterior estimation with swyft: stop wasting your precious time. NeurIPS workshop on Machine Learning and the Physical Sciences. arXiv:2011.13951. 11 Dec 2020.

Benjamin Kurt Miller, Mario Geiger, Tess E. Smidt, and Frank Noe. Relevance of rotationally equivariant convolutions for predicting molecular properties. NeurIPS workshop on Interpretable Inductive Biases and Physically Structured Learning. arXiv:2008.08461. 12 Dec 2020. (and Machine Learning for Molecules Workshop 2020)

Tess E. Smidt, Mario Geiger, Benjamin Kurt Miller. Finding Symmetry Breaking Order Parameters with Euclidean Neural Networks. arXiv: 2007.02005. 7 July 2020.

Benjamin Kurt Miller. SE(3) Equivariant Neural Networks for Regression on Molecular Properties: The QM9 Benchmark. Refubium Freie Universität Berlin Repository. 31 March 2020. Berlin, Germany.

Benjamin K. Miller. Proper Orthogonal Decomposition (Download pdf). Coursework at Freie Universitaet Berlin. 14 November 2018. Berlin, Germany.

Benjamin K. Miller. Classical Analysis of High Harmonic Generation. University of Colorado Boulder Honors Thesis. 5 November 2015. Boulder, Colorado, USA.

* denotes equal contribution or core contributors.

Talking

A Few Generative Models for Crystals. AI-Accelerated Superconductor Discovery: From Algorithms to Novel Materials. 29 Apr 2025. Gainesville, Florida, USA.

FlowMM and FlowLLM. Microsoft Research Amsterdam. 21 Nov 2024. Amsterdam, the Netherlands (online).

FlowMM (and more). Machine Learning and Informatics for Chemistry and Materials. Telluride Science and Innovation Center. 17 Oct 2024. Telluride, Colorado, USA.

FlowMM: Generating Materials with Riemannian Flow Matching. Christopher Sutton Lab. 2 Oct 2024. Online.

Ratios and more. SBI Hackathon. 21 Mar 2024. Tuebingen, Germany.

An overview of Simulation-based Inference. AI4Science Symposium. 24 Feb 2023. Amsterdam, the Netherlands.

An overview of Simulation-based Inference methods and software. Simulation-based inference with swyft. Jan 23rd - 26th 2023. Amsterdam, the Netherlands.

Truncated Marginal Neural Ratio Estimation with swyft. 5th Inter-experiment Machine Learning Workshop. May 9th - 13th 2022. CERN, Meyrin, Switzerland.

Truncated Marginal Neural Ratio Estimation with swyft. Likelihood-free in Paris. 21 April 2022. École Normale Supérieure, Paris, France.

Truncated Marginal Neural Ratio Estimation. NeurIPS 2021 Ghent Meetup. 8 Dec 2021. Ghent, Belgium. (Canceled due to COVID-19)

Benjamin Kurt Miller, Patrick Forré, Christoph Weniger. Marginal Posteriors with Truncated Marginal Ratio Estimation. SIAM AN21. CP17 Probability and Statistics. 23 July 2021. Virtual.

Constrained marginal likelihood-to-evidence ratio estimation. AI4Science Colloquium. 18 March 2021. Amsterdam, the Netherlands.

Constrained Marginal Likelihood-to-evidence ratio estimation: projections of the posterior distribution. Atlas Experiment at CERN Statistics and Machine Learning Colloquium. 18 March 2021. Online.

Simulation-based Marginal Inference. EuroMoonMars, Earth, Space & Innovation EMMESI Workshop. 16-19 March 2021. Virtual Hybrid Workshop.

Benjamin Kurt Miller, Alex Cole, Gilles Louppe, and Christoph Weniger. Simulation-efficient marginal posterior estimation with swyft. Gravitational waves: a new messenger to explore the universe. 1 March 2021. Digital Conference.

Determining astrophysical parameters with machine learning. AI4Science Kickoff Workshop. 9 July 2020. Amsterdam, the Netherlands.

Benjamin K. Miller, Tess E. Smidt, Mario Geiger. Consequences of Symmetry on Articulating Machine Learning Tasks. Society for Industrial and Applied Mathematics Conference on Mathematics of Data Science. 5-7 May 2020. Cincinnati, Ohio, USA. (Canceled due to COVID-19)

Benjamin K. Miller, Tess E. Smidt, James A. Sethian. Applications of SE(3) Equivariant Neural Networks. Berkeley Lab Computing Science Summer Program Poster Session. 1 August 2019. Berkeley, California, USA.

Benjamin Kurt Miller, Christoph Redlich, Lina Jaurigue, Benjamin Lingnau, and Kathy Lüdge. Gain compression induced polarization mode competition in quantum-dot micropillar lasers: Effects of coherent feedback on multi-mode rate equations. Deutsche Physikalische Gesellschaft. 20 March 2017. Dresden, Germany.

Benjamin K. Miller and Agnieszka Jaron-Becker. Classically Modeling High Harmonic Generation with a focus on Elliptical Polarization. JILA Posterfest. 2 October 2015. Boulder, Colorado, USA.

Bold denotes invited talk.

Code

Christoph Weniger, Benjamin Kurt Miller, Francesco Nattino, Alex Cole, Ou Ku, Meiert Willem Grootes, & Adam Coogan. (2021). undark-lab/swyft. Zenodo. https://doi.org/10.5281/zenodo.5752734.

Mario Geiger, Tess Smidt, Alby M., Benjamin Kurt Miller, et. al. (2021). e3nn/e3nn. Zenodo. https://doi.org/10.5281/zenodo.3724963.

Memberships

ELLIS PhD Program - University of Amsterdam Max Welling - University of Liège Gilles Louppe

Students

Wenkai Pan - University of Amsterdam MSc in Artificial Intelligence
Nienke Duetz - University of Amsterdam MSc in Artificial Intelligence
Christoph Hönes - University of Amsterdam MSc in Artificial Intelligence

Review

NeurIPS 2023
ICML 2023
ICLR Workshop Physics4ML 2023
NeurIPS 2022
ICML 2022
NeurIPS Machine Learning for Molecules Workshop 2021
NeurIPS Workshop on Machine Learning and the Physical Sciences 2021

Research Visits

Gilles Louppe at University of Liège in Belgium. Feb 13 - 17.
Gilles Louppe at University of Liège in Belgium. Jan 16 - 20.

Mathematics Education Resources

I do not contribute to these sources but I am a fan. For that reason, I will share them here.

Richard E. Borcherds
Tensors for Beginners by eigenchris
Math Doctor Bob
MIT OpenCourseWare Probabilistic Systems Analysis and Applied Probability
The Bright Side of Mathematics