The IG 'Machine Learning' (ML@NFDI4Earth) within the framework of NFDI4Earth addresses the growing digital needs of the individual disciplines and sub-disciplines within all addressed Earth system sciences (ESS). The aim of ML@NFDI4Earth is to maximize synergies and bundle information about ML activities in the field of ESS and to improve exchange between stakeholders across disciplinary boundaries.
Summary, i.e. ML@NFDI4EARTH Topics
The network will gather and identify projects in the area of ML@NFDI4Earth, research topics, used techniques, best practices, software, frameworks, benchmarks, data sets, hardware setups, etc.
Concrete questions are:
- How are ML frameworks installed at RDIs?
- How to interface common ML frameworks with data stored in the ESS?
- Have training, validation, test and verification data sets enough meta data?
- Do data management plans consider ML data sets already?
- Scalability of approaches from laptop to HPC?
- Are the interfaces available to fully exploit new ML in rather old Earth system modeling and analysis frameworks? (e.g. python / fortran / Julia interfaces)
Illustrating material & further links
Machine Learning at NFDI4Earth (slides of 1st Plenary Meeting 2022)
Mailing list to subsrice
Current agenda and next meetings resp.