NOMAD Laboratory
NOMAD Centre of Excellence

Bringing computational materials science to exascale

The NOMAD CoE aims at assessing and exploiting the characteristics of next generation HPC for ab initio computational materials science, to enable investigations of systems of higher complexity (space and time), consideration of metastable states and temperature, and all this at significantly higher accuracy and precision than what is possible today.

Exascale Codes

  • Bringing DFT, Green-function methods, and coupled-cluster theory to exascale
  • Supporting entire code families, covering planewaves (PW), linearized augmented PWs, and atom-centred orbitals
  • Follow us on GitHub

Exascale Workflows

  • Enabling exascale computations by advanced workflows
  • Covering high-throughput computations and beyond-DFT workflows
  • Learn how to work with ASE/ASR and FireWorks in this tutorial

Extreme-scale data

  • Advance the NOMAD AI toolkit and bring it towards near-real-time performance
  • Like to visit the NOMAD Laboratory and its services for up- and downloading, and exploring materials data? 
  • Watch our video tutorials to learn how to work with the AI toolkit

Please see our news feed on Twitter and/or subscribe.

Tweets by NoMaDCoE
Nov 22, 2021

New Website for the NOMAD AI-Toolkit

New Website for the NOMAD AI-Toolkit

The NOMAD Artificial-Intelligence (AI) Toolkit has been established in order to create a shared environment where researchers belonging to the materials-science community are invited to develop and re-use artificial-intelligence-driven projects, as well as learn or (further) develop AI methods.

The AI Toolkit can be easily operated via a web browser and provides access to Jupyter notebooks, which are interactive web pages for editing and executing code. Through the NOMAD API, users have direct access to data in the NOMAD Archive & Repository  and can train or apply AI algorithms, including data mining and machine learning. Algorithms such as deep and/or convolutional neural networks, kernel ridge regression, decision trees, as well as algorithms recently introduced by the NOMAD team such as SISSO and subgroup discovery are part of the AI Toolkit's extensive repertoire. Users can interactively train, optimize and test AI models and visualize the result in publication-quality graphics. Tutorials are ranked by difficulty level, so that users from newcomers to experts can find interesting and challenging material to improve their skill set in the field.

The AI Toolkit is motivated by and brings to the next level the FAIR principles for scientific research. FAIR data are Findable, Accessible, Interoperable and Reusable (or Recyclable). The AI Toolkit adds to these aspects the FAIRness of the AI algorithms used for the analysis of the data. In other words, users can peruse the workflow of published works, from the data retrieval and pre-processing through the training of the AI models, to the production of the resulting plots and tables, all in an interactive fashion, i.e., being able to modify the workflow and possibly re-use it.

In this sense, the AI toolkit reinterprets the AIR in the FAIR acronym as “AI-ready data”.

By using the AI toolkit, users from both academia and industry are enabled to learn, by putting their hands-on, all the details of successful research combining AI and materials science, in order to learn new techniques and  promptly apply the to their scientific topics of interests.

A new review article by Luca Ghiringhelli (NOMAD CoE), published in the Research Highlights section of the prestigious journal Nature Reviews Physics in September 2021, describes the capabilities of the NOMAD Artificial Intelligence (AI) Toolkit for discovering new materials with exceptional performance or for uncovering new properties of existing materials.

L. M. Ghiringhelli: An AI-toolkit to develop and share research into new materials. Nature Review Physics 3, 724 (2021);