Machine Learning Potentials: from interfaces to solution

A 5-Day Hands-on Tutorial Workshop

 

We are happy to announce “MLPfits 2027”, a 5-day hands-on Tutorial Workshop on training and using machine learning potentials for large-scale atomistic simulations in chemistry and materials science.

Dates: February 15-19, 2027 (start: February 15, 8:30 am, end: February 19, noon)

Place: Ruhr-Universität Bochum, Germany (onsite only)

Scope of the Workshop

In recent years, Machine Learning Potentials (MLPs) have attracted a lot of attention and many different methods have been proposed to date. By training on datasets derived from accurate quantum mechanical calculations, MLPs can approximate the complex multidimensional potential energy surface with ab initio accuracy while yielding energies and forces at drastically reduced computational costs enabling large-scale simulations.

The aim of the workshop is to familiarize researchers with the background and application of two important families of machine learning potentials:

1. HDNNP: We will cover different generations of high-dimensional neural network potentials (HDNNP) based on the Behler Parrinello method. Next to efficient second-generation HDNNPs, also HDNNPs of the third and fourth generation including long-range interactions in their new implementation in the RuNNer2 code will be covered.

2. ACE and GRACE: We will present the Atomic Cluster Expansion (ACE) and its generalization to graph basis functions (GRACE). ACE provides a rigorous mathematical foundation that encompasses and connects various other models, including equivariant message passing neural networks, while allowing for extensions to diverse atomic properties. Participants will train GRACE models, explore their versatile applications, and learn strategies for fine-tuning foundational GRACE potentials to tackle specialized research problems.

Further, we will discuss strategies for the validation of machine learning interatomic potentials.

Registration:

The registration for the workshop will open soon.

Workshop Program:

The program of the workshop will be announced in early 2027. It will consist of tutorial lectures and hands on  computer sessions.

 

Contact:

E-mail: potentials@icams.rub.de

 

We are looking forward to welcoming you to Bochum,

Jörg Behler and Ralf Drautz

 

 

We are grateful to the Sponsors of the Workshop:

Funding has been provided by:

            

Further, we are grateful to Paderborn Center for Parallel Computing for kindly providing the IT infrastructure for the workshop:

         

 

 

Previous Organized Meetings

Workshop MLPfits 2026
"Machine Learning Potentials: from interfaces to solution"
(with Ralf Drautz)
Bochum, 2.-5. March 2026

CECAM Workshop
"A roadmap for an atomistic machine learning software ecosystem"
(with Michele Ceriotti, Cecilia Clementi and Gábor Csányi)
Lausanne, 19.-21. January 2026

CECAM Workshop
"Machine Learning Potentials: from interfaces to solution (MLPfits)"
(with Amin Alibakhshi, Ralf Drautz and Lars Schäfer)
Ruhr-Universität Bochum, 27.-29. August 2024

Workshop
"From Electrons to Phase Diagrams"
(with Ralf Drautz, Jörg Neugebauer, Karsten Albe)
Ruhr-Universität Bochum, 8.-10. June 2022

ICASEC Summer School
"Virtual Summer School on Theoretical Methods for Energy Conversion"
(with Ricardo Mata, Alec Wodtke)
Göttingen, 14.-16. June 2021

Workshop
"Developing High-Dimensional Potential Energy Surfaces – From the Gas Phase to Materials"
(with Joel Bowman, Sascha Kandratsenka)
Göttingen, April 24-26, 2019

Conference
"52nd Symposium for Theoretical Chemistry"
(with Dominik Marx, Christof Hättig, Michael Römelt, Lars Schäfer)
Ruhr-Universität Bochum, 26.-29. September 2016

Workshop
"Machine Learning in Atomistic Simulations"
(with Gareth Tribello)
Lugano, September 10-12, 2012

Workshop
"Non-Adiabatic Dynamics at Surfaces"
(with Patrick Rinke, Karsten Reuter, Matthias Scheffler)
Schloss Reisensburg, October 22-25, 2007