Liverpool Texture Hero Banner

Machine Learning for materials science: community requirements for building a UK capacity

Royce Partner the University of Liverpool recently convened a one-day workshop to canvas a wide range of materials’ science practitioners who have been early adopters of the use of machine learning.

The group was made up of representatives from academia, industry and the public sector, working in experimental and computational materials science, and spanning different career stages.

The workshop explored the challenges faced in applying ML for materials science under five broad categories:

  • Education/Training
  • Infrastructure
  • Ethics
  • Open Science and FAIR data
  • Community building and events.

The focused discussions and round-table format allowed formulation of key requirements for building a world-leading machine learning for materials science research capacity in the UK.

Across the five themes outlined above, it was agreed that there is a need for the following:

  • Establishing regular, sustainably funded training specifically for practitioners of ML in materials science
  • A significant increase in the compute capacity for ML
  • A dedicated, sustainably funded infrastructure for data storage and data and model sharing
  • Establishing of an ethics framework around ML research in materials science
  • A requirement for environmental impact statements in ML research proposals
  • Explicit consideration of contributions to Open Science in research assessment exercises for all publicly funded organisations
  • Establishing community-led, national-level bodies to act as a focal point of the ML in materials science community, in the form of learned body interest groups, AI hubs and/or an ML in Materials Science network Training/Education

As a short-term outcome of the workshop, a special interest group for ML in materials science will be established. This group will be hosted across learned bodies, specifically the Institute of Physics (IOP) and the Royal Society of Chemistry (RSC), reflecting the make-up of the community at the meeting.  It was also agreed that there is a need to develop regular events to provide a focal point for the community.

The full report can be downloaded here