🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Train one Machine-Learning Potential with another! MLPs capture interactions between atoms with DFT accuracy at a fraction of the cost. Combining MLPs reduces this further, making million-atom simulations feasible doi.org/10.1063/5.0099… #RSCDigital #RSCMat #RSCPhys #RSCPoster

👋 to everyone at the #RSCPoster Conference! I'm very pleased to present my work combining synthetic data and atomistic machine learning. #RSCDigital #RSCPhys #RSCMat #compchem (preprint at arxiv.org/abs/2211.16443)

Excited to show you a complete first-principles stability range of phosphorus modifications! Check our preprint (arxiv.org/abs/2211.04771) from Deringer Group Oxford on the chemical bonding of amorphous phosphorus (a-P) using ML and DFT methods. #RSCPoster #RSCInorg #RSCPhys #compchem

Here's our Deringer Group Oxford #RSCPoster contributions - please do visit & follow! 🙂 ➡️ John (John Gardner) on synthetic data for atomistic ML ➡️ Joe (Joe Morrow) on teacher-student ML potentials ➡️ Zak (Zakariya El-Machachi) on amorphous graphene ➡️ Yuxing (Yuxing Zhou) on red phosphorus

Online now at NatRevMaterials! Very happy to have contributed to this Viewpoint on future directions in materials simulations 😀🤖 #compchem nature.com/articles/s4157…

Congratulations to all #RSCPoster prize winners - including our own Zakariya El-Machachi who received a runner-up prize in the #RSCDigital category! 🎉 Read more about Zak's #compchem work in Chemical Science: pubs.rsc.org/en/content/art…

A good example of combining machine-learned potentials with chemical bonding analysis (using LOBSTER 🦞!) - both are my favorite tools to study inorganic chemistry 😆! Special thanks to Stephen and Volker Deringer for fabulous contributions!

Very excited that my Master’s project has made it into preprint form!🥳 Read all about it here: arxiv.org/abs/2305.05536 Thank you to my supervisors Volker Deringer, Andrew Goodwin and to my co-authors @thomascnicholas and John Gardner for all their hard work and support. (1/3)

New group photo! 😀 Such a pleasure to see team Deringer Group Oxford expanding and going strong - thanks to everyone's creative contributions, & to generous support by UK Research and Innovation Engineering and Physical Sciences Research Council and of course Oxford Chemistry

Excited to be here in Dublin! We will show how machine learning-based methods can be used to study very disordered and complex (two-dimensional) material systems, and to enable device-scale atomistic simulations, both on quantum mechanical accuracy. Come and have a talk! #MC16

Hello Dublin 😀 Excited to be at Royal Society of Chemistry #MC16 where talks on materials chemistry (& #compchem materials design) have already started!

Exciting news! Fernanda Duarte and I will be jointly chairing an Royal Society of Chemistry Faraday Discussions meeting on "data-driven discovery" in chemistry. We aim to bring together the molecular & materials communities to discuss databases, AI/ML, and more. See rsc.li/3QdkKDz (+ thread) (1/3)

Slightly delayed (!!) but very pleased to announce that our paper: "Synthetic data for pre-training neural-network potentials" is out in Machine Learning: Science and Technology See 👇🧵 for a quick summary, or have a complete read here: doi.org/10.1088/2632-2… All thoughts and comments very welcome! 1/

A super helpful tool to use various (popular) training databases of atomic structures for #ML purposes. It covers a wide range of organic and inorganic systems, including our recently published training database for Ge-Sb-Te alloys! Huge congrats to John Gardner’s great work!

Looks cool!

Want to use ACE but don’t have sufficient experience with the hyperparameters? XPOT might be a good starting point! Check our latest preprint on hyperparameter optimisation using XPOT. Other interesting works will follow 🙌🏻

Many many thanks to Daniel Daniel Thomas du Toit and Volker Volker Deringer for their great efforts. Feel free to talk to Daniel about more technical details!

Check the useful graph-pes from John Gardner for fitting graph-based ML potentials!

I've been working hard behind the scenes, and am excited to announce that graph-pes version 0.1.0 is now out!