Skip to main content

Data to knowledge

This webinar investigates machine learning for interatomic potentials (MLIP).

About the Webinar​

Welcome to the Physical Sciences Data Infrastructure (PSDI) webinar series. This webinar series is designed to communicate the PSDI work to a wider audience!

The subject of this webinar, held on 14th December 2023, is our pathfinder focused on transforming data to knowledge through the construction of workflows. In particular this looks at machine learning for interatomic potentials (MLIP). This webinar was presented by Alin Elena and Federica Zanca from Science and Technology Facilities Council Daresbury Laboratory.

Abstract​

AI is ubiquitous in all walks of life and research. In the last 5 years active research in the field of understanding how atoms and molecules interact using AI has revolutionised the field. The outcome of these efforts, machine learning interatomic potentials, MLIP, has produced the breakthrough that recommends them as the next paradigm change in atomistic molecular simulations, with applications ranging from battery design to catalytic chemical reaction modelling for hydrogen storage or CO2 capture. All these advances need expensive calculations to be produced and used for training machine learning models. In addition, the models resulting from these models are non-trivial in terms of storage and distribution compared with previous generation interatomic potentials which tended to be analytical. Current models use millions of structures for training, and this raises new challenges around reproducibility and exploitation of these models. Physical Sciences Data Infrastructure, PSDI, aims to enable researchers in the field to deal with the challenges coming from generating, using and enhancing this data. The first part of the talk will concentrate on recent advances in the field and PSDI contribution, second part more technical will concentrate on presenting databases and workflows challenges for having reproducible and reusable data. We will present open search, abcd database and aiida workflows.

Watch the recording​

You can watch this recording via our You Tube channel.

What to do next​

Related links:

  • Galaxy Training
  • Elixir TeSS: extensive training materials with a focus on computation in the life sciences, but many courses are also relevant for the physical sciences community.
About this page

If you would like to contribute content to the PSDI Knowledge Base or have feedback you would like to give on this guidance, please contact us.