Ben Moseley

Our Scalable Scientific Machine Learning Lab @ Imperial College London website is now live! We build robust, scalable SciML methods to accelerate science — from de-noising Moon images to speeding up multi-scale simulation @ESEImperial @imperialcollege scalable-sciml-lab.org

We just released our ETH Zürich AI in the Sciences and Engineering Master’s course on YouTube! 📚 Prof. Siddhartha Mishra and I explain PINNs, neural operators, foundation models, neural PDEs, diffusion models, symbolic regression, and more! #AI4Science youtube.com/watch?v=LkKvhv…

Can we carry out multi-scale simulation with physics-informed neural networks (PINNs)? PINNs often struggle to solve multi-scale problems, and in our new paper we take steps to overcome this by combining them with multiple levels of domain decompositions: doi.org/10.1016/j.cma.…

Just released a major new update to our FBPINNs library, which now uses JAX. Solve forward and inverse problems related to PDEs using PINNs + domain decomposition, and scale to 1000s+ subdomains! #jax #pinns #PDEs github.com/benmoseley/FBP…

Very excited to announce that our entire ETH Zürich Deep Learning in Scientific Computing Master's course is now on YouTube! 📖 Prof. Siddhartha Mishra and I will talk you through PINNs, neural operators, neural ODEs, differentiable physics and more youtube.com/watch?v=y6wHpR…

I will be giving a mini-course on efficient physics-informed neural networks at the CWI Scientific Machine Learning and Numerical Methods Autumn School in October in the Netherlands - open to everyone, more info here: cwi.nl/en/events/cwi-…

Faster population balance equation (PBE) simulations with JAX! PBEs model crystallization, chemical reactors & biological systems - but traditional solvers are too slow. Our new JAX solver accelerates computation & enables AI-driven physics discovery. pubs.acs.org/doi/10.1021/ac…

New paper: accurately modelling multiphase fluid flow in fractured porous media using physics-informed neural networks, validated using real CT-scan data doi.org/10.1016/j.cma.…. Work led by Jassem Abbasi and others - it was a pleasure to collaborate!

Having recently joined the department, I can say I have felt very supported and inspired starting my lab here, and it would be great to work alongside more experts in these areas! @ESEImperial

💫Fellow scientific computing geeks, here is an example of adaptive mesh refinement in #JAX: gist.github.com/patrick-kidger… I recently heard a claim that this was impossible in JAX due to its lack of dynamic shapes, and decided to prove the naysayers wrong 😄

🚀 Excited to share that I’m joining @ImperialCollege in Nov as a Lecturer, leading the Scalable SciML Lab in @ESEImperial! I’m also honoured to be a Schmidt AI in Science Fellow at @ImperialX_AI. I'm hiring PhD students! benmoseley.blog 🌍✨

This may be the most important figure in LLM research since the OG Chinchilla scaling law in 2022. The key insight is 2 curves working in tandem. Not one. People have been predicting a stagnation in LLM capability by extrapolating the training scaling law, yet they didn't…

🚀 Exciting news! Applications for ETH AI Center PhD and Post-Doc Fellowships are now OPEN! 🤩 Apply by Nov 22, 2023, and dive into the world of #AI, #MachineLearning, #DataScience, and more. Details at: ai.ethz.ch/apply #PhD #PostDoc #AIresearch #ETHZ

Here is the recording: video.ethz.ch/events/2023/ph…

I will be giving a guest lecture for the Physical Society of Zürich tonight 7.30pm on "How to incorporate physical understanding into machine learning" - admission is free and the talk will be recorded afterwards! pgz.ch/events/ws2324/…

Interested in investigating how elephants use ground vibrations to communicate? + collecting data in Kenya? The Animal Vibration Lab at the @UniofOxford (a group I have collaborated with) has an open postdoctoral position! tinyurl.com/5xucd5fe

Excited to announce that I am interning in NVIDIA's Modulus team. Modulus is a scientific machine learning framework which blends together neural networks with PDEs, allowing more robust models to be built for scientific analysis: developer.nvidia.com/modulus

I will be presenting our latest research on finite basis physics-informed neural networks as a poster at The Symbiosis of Deep Learning and Differential Equations workshop at NeurIPS tomorrow: come along to find out more! dl-de.github.io arxiv.org/abs/2107.07871

Ben Moseley @benm_oseley, a DPhil student co-supervised by Andrew Markham @CompSciOxford and Tarje Nissen-Meyer @OxUniEarthSci wrote a blog-post that has had over 1M views on Linkedin. Read ‘So, what is a physics-informed neural network?’ here: bit.ly/2ZGOFxF