About

About me

I am currently a PhD student working on physics and machine learning at the University of Oxford’s Center for Autonomous Intelligent Machines and Systems (2017-present), a machine learning lead for NASA’s Frontier Development Lab (2019-present), and a scientific machine learning intern at NVIDIA (2022-present).

Prior to this I was a geophysicist at BP (2012-2017), and studied physics (masters degree) at Durham University, UK (2008-2012).

GitHub
LinkedIn
Twitter
Google Scholar
Mail

Research interests

I am interested in machine learning for physics. More specifically, I want to understand how best we can combine physical principles and machine learning together in more nuanced ways, which will hopefully lead to more powerful tools for AI researchers and other scientists alike. In my work I have developed new physics-informed machine learning algorithms for solving differential equations and applied state-of-the-art physics-based machine learning to many different real-world scientific problems, ranging from searching for water on the Moon to tracking elephants in Kenya.

I also want to inform the world about AI. I do this by writing AI blogs, teaching, releasing open-source code and publishing research on the intersection of AI and physics.

News

  • [Feb 22] I was a course tutor and lead teaching assistant for the University of Oxford Department of Computer Science undergraduate courses in Artificial Intelligence and Advanced Topics in Machine Learning.
  • [Dec 21] I will be interning half-time within NVIDIA’s Modulus team, contributing to their state-of-the-art scientific machine learning framework.
  • [Nov 21] two NeurIPS workshop papers accepted (The Symbiosis of Deep Learning and Differential Equations and Machine Learning and the Physical Sciences)
  • [Oct 21] My blog post on physics-informed neural networks hit 1M+ views! Thank you for all of the great conversation on this post!
  • [Sep 21] our Nature Communications paper on peering into permanently shadowed regions on the Moon’s surface with machine learning is out. Read the press coverage from the Max Planck Institute and University of Oxford here and here.
  • [Jul 21] our latest preprint (and code) is out, which proposes a scalable method for solving differential equations using physics-informed neural networks (FBPINNs).
  • [Jul 21] our paper using ground vibrations to localise elephants has been published in the Journal of The Royal Society Interface. Read the press coverage from the University of Oxford here.
  • [Jun 21] I will be returning to NASA’s Frontier Development Lab this year as a faculty member (ML team lead), helping to super-resolve images of the lunar surface.
  • [Mar 21] our paper on denoising extremely low-light images of the lunar surface has been accepted at CVPR.
  • [Feb 21] I was a teaching assistant for the University of Oxford Department of Computer Science undergraduate course in Advanced Topics in Machine Learning.
  • [Nov 20] I was a teaching assistant for the University of Oxford Centre for Autonomous Intelligent Machines and Systems postgraduate course on Machine Learning.
  • [Sep 20] I helped to accelerate wind energy production in the UK by leading a team of 8 ML researchers at the UK Alan Turing Institute 2020 September Data Study Group.
  • [Aug 20] Our manifesto paper on using deep learning to accelerate seismic simulation has been published in Solid Earth.
  • [Jul 20] Our paper using unsupervised machine learning to discover thermal anomalies on the Moon has been published in the Planetary Science Journal.
  • [Jun 20] our latest preprint is out, on using physics-informed neural networks to solve the wave equation.
  • [Feb 20] our team won best technical solution at the Oxford Foundry AI & Climate Change Hackathon.
  • [Jan 20] I was invited to be a Series Editor for Elsevier’s 2020 Advances in Geophysics special issue on Machine Learning in the Geosciences.
  • [Jun 19] I will be joining NASA’s Frontier Development lab over this summer as a AI researcher, working in their Moon For Good team.

Invited talks

[2022] Scientific machine learning: ways to incorporate scientific principles into ML (Roche digital day / Lucerne University / University of Wyoming)
[2022] Finite Basis Physics-Informed Neural Networks (FBPINNs): a scalable domain
decomposition approach for solving differential equations (link) (Brown University CRUNCH group)
[2021] Physics-informed deep learning for solving differential equations (NVIDIA research / UK Science and Technology Facilities Council)
[2021] CVPR main conference poster presentation
[2021] Enhancing images of lunar permanently shadowed regions with machine learning (NASA/VIPER)
[2021] Unsupervised learning for lunar exploration (UKRI Joint CDT Conference in Robotics and AI)
[2020] NeurIPS Machine Learning and the Physical Sciences workshop poster presentation
[2020] Solving the wave equation with physics-informed ML (Stanford Earth seminar)
[2020] Anomaly detection with variational autoencoders (ETH Zurich Machine Learning and Astrophysics workshop / UCLA Diviner experiment research seminar)

Recent publications

See full publication list

Key: plain language summary

Bickel, V., Moseley, B., Lopez-Francos, I., Shirley, M. (2021). Peering into Lunar Permanently Shadowed Regions with Deep Learning. Nature Communications. [press coverage] Used machine learning to enhance images of permanently shadowed regions on the Moon, allowing us to see features within these regions down to 3 m in size for the first time.

Moseley, B., Bickel, V., Lopez-Francos, I. G., & Rana, L. (2021). Extreme Low-Light Environment-Driven Image Denoising Over Permanently Shadowed Lunar Regions With a Physical Noise Model. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). [blog post] Used a combination of physical modelling and deep learning to remove noise from images of permanently shadowed regions on the lunar surface, aiding the search for water and future mission planners.

Moseley, B., Markham, A., & Nissen-Meyer, T. (2021). Finite Basis Physics-Informed Neural Networks (FBPINNs): a scalable domain decomposition approach for solving differential equations. ArXiv. [code] Proposed a new method for solving differential equations which is able to scale to large problems by using physics-informed neural networks and a divide-and-conquer strategy.

Reinwald, M., Moseley, B., Szenicer, A., Nissen-Meyer, T., Oduor, S., Vollrath, F., … Mortimer, B. (2021). Seismic localization of elephant rumbles as a monitoring approach. Journal of The Royal Society Interface. [press coverage] [blog post] [code] Showed that the location of elephants can be accurately determined by using the ground vibrations they make whilst calling each other and a positioning algorithm which is able to say how uncertain it is.

Moseley, B., Markham, A., & Nissen-Meyer, T. (2020). Solving the wave equation with physics-informed deep learning. ArXiv. Showed that physics-informed neural networks can be used to simulate complex wave physics, and that they potentially offer a faster approach than traditional simulation methods.

Moseley, B., Nissen-Meyer, T., & Markham, A. (2020). Deep learning for fast simulation of seismic waves in complex media. Solid Earth. [code] Discussed in detail the advantages and challenges of using deep learning for simulating seismic waves, and presented multiple deep neural networks which can significantly accelerate traditional methods.

Moseley, B., Bickel, V., Burelbach, J., & Relatores, N. (2020). Unsupervised Learning for Thermophysical Analysis on the Lunar Surface. The Planetary Science Journal. [blog post] Used machine learning to disentangle the underlying factors of variation in the Moon’s surface temperature, allowing us to identify thermal anomalies on the surface and verify existing physical models.

Latest blog posts

See all blog posts

Latest comments