The NOMAD Artificial-Intelligence Toolkit: Web-Based FAIR-Data-Driven Materials Science

The Novel-Materials Discovery (NOMAD) Laboratory created and maintains the Repository & Archive, the largest data store of computational materials data worldwide, which stores more than 100 million calculations. Here, we present the NOMAD Artificial-Intelligence (AI) Toolkit, a web-based infrastructure for the interactive analysis of the material-science Findable, Accessible, Interoperable, and Recyclable (FAIR) data stored in the NOMAD Archive. By using Jupyter notebooks running in a web-browser (no software to be installed on the user side), the NOMAD data can be accessed and data mining, machine learning, and other AI techniques can be applied to analyze them. This infrastructure brings the concept of reproducibility in materials science to the next level, by allowing researchers to share, besides the data contributing to their scientific publications, also all the analytics tools they have created, adapted, and applied for unveiling patterns in them and predicting properties of known, new, or even novel materials.
The Jupyter notebooks, all reachable via https://nomad-coe.eu/AIToolkit, span interactive tutorials reproducing the full AI workflow of recent landmark publications and shallow-learning-curve tutorials on textbook as well as recently developed AI techniques.